Time Series Good

  • May 2020
  • PDF

This document was uploaded by user and they confirmed that they have the permission to share it. If you are author or own the copyright of this book, please report to us by using this DMCA report form. Report DMCA


Overview

Download & View Time Series Good as PDF for free.

More details

  • Words: 55,889
  • Pages: 204
‫ﺟﺎﻣﻌﺔ اﻟﻤﻠﻚ ﺳﻌﻮد‬ ‫آﻠﻴﺔ اﻟﻌﻠﻮم‬ ‫ﻗﺴﻢ اﻹﺣﺼﺎء وﺑﺤﻮث اﻟﻌﻤﻠﻴﺎت‬

‫ﻃﺮق اﻟﺘﻨﺒﺆ اﻹﺣﺼﺎﺋﻲ‬ ‫) اﻟﺠﺰء اﻷول(‬

‫ﺗﺄﻟﻴﻒ د‪ .‬ﻋﺪﻧﺎن ﻣﺎﺟﺪ ﻋﺒﺪاﻟﺮﺣﻤﻦ ﺑﺮي‬ ‫أﺳﺘﺎذ اﻹﺣﺼﺎء وﺑﺤﻮث اﻟﻌﻤﻠﻴﺎت اﻟﻤﺸﺎرك‬

٢

‫ﺑﺴﻢ اﷲ اﻟﺮﺣﻤﻦ اﻟﺮﺣﻴﻢ‬ ‫اﻟﺤﻤﺪ ﷲ رب اﻟﻌﺎﻟﻤﻴﻦ واﻟﺼﻼة واﻟﺴﻼم ﻋﻠﻰ اﺷﺮف ﺧﻠﻖ اﷲ ﺳﻴﺪﻧﺎ وﻧﺒﻴﻨﺎ ﻣﺤﻤﺪ‬ ‫وﻋﻠﻰ ﺁﻟﻪ وﺻﺤﺒﻪ وﺳﻠﻢ‪.‬‬ ‫أﻣﺎ ﺑﻌﺪ‪.‬‬ ‫هﺬﻩ هﻲ اﻟﻤﺴﻮدة اﻷوﻟﻰ ﻟﻜﺘﺎب ﻃﺮق اﻟﺘﻨﺒﺆ اﻹﺣﺼﺎﺋﻲ ﻟﻄﻼب ﻣﺮﺣﻠﺔ اﻟﺒﻜﺎﻟﻮرﻳﻮس‪.‬‬ ‫هﺬا اﻟﻜﺘﺎب ﺳﻴﻈﻞ ﻣﺴﻮدة إﻟﻰ ﻣﺎﺷﺎء اﷲ ﻷﻧﻲ وﺑﺈذن اﷲ ﺗﻌﺎﻟﻰ ﺳﻮف أﻗﻮم ﺑﺘﻄﻮﻳﺮﻩ‬ ‫وﺗﺠﺪﻳﺪﻩ وﺗﺤﺴﻴﻨﻪ ﺑﺸﻜﻞ ﻣﺴﺘﻤﺮ وﺳﻴﻈﻞ ﺑﺸﻜﻠﻪ اﻹﻟﻜﺘﺮوﻧﻲ هﺬا ﻷﻧﻲ أﻋﺘﻘﺪ ان اﻟﻌﻠﻮم‬ ‫واﻟﺘﻘﻨﻴﺔ ﺗﺘﻄﻮر ﻳﻮﻣﻴﺎ وﺑﺸﻜﻞ ﻣﺘﺴﺎرع ﺑﺤﻴﺚ ان وﺿﻌﻬﺎ ﻓﻲ آﺘﺎب ﺟﺎﻣﺪ ﺳﺘﺎﺗﻴﻜﻲ‬ ‫ﻻﻳﺘﻨﺎﺳﺐ ﻣﻊ دﻳﻨﺎﻣﻴﻜﻴﺔ اﻟﻤﻮﺿﻮع وﺧﺎﺻﺔ ﻓﻲ ﻋﺼﺮ ﺛﻮرة اﻟﻤﻌﻠﻮﻣﺎت واﻹﻧﺘﺮﻧﺖ‪.‬‬ ‫ﻳﻐﻄﻲ اﻟﺠﺰء اﻷول ﻣﻦ اﻟﻜﺘﺎب اﻷﺳﺎﺳﻴﺎت اﻷوﻟﻴﺔ ﻟﻠﻤﻮﺿﻮع وﻳﺘﻄﺮق إﻟﻰ ﻣﻮﺿﻮع‬ ‫اﻟﺘﻨﺒﺆ اﻹﺣﺼﺎﺋﻲ ﺑﺈﺳﺘﺨﺪام ﻧﻤﺎذج ‪ ARIMA‬واﻟﺘﻲ آﺎﻧﺖ اول ﻣﻌﺎﻟﺠﺔ رﻳﺎﺿﻴﺔ ﺟﺎدة‬ ‫وﻣﺤﻜﻤﺔ ﻟﻠﺘﻨﺒﺆ اﻹﺣﺼﺎﺋﻲ ﺑﺈﺳﺘﺨﺪام اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ ‪ Time Series‬آﻤﺎ ﺗﻄﺮﻗﺖ‬ ‫ﻓﻲ ﺁﺧﺮ اﻟﻜﺘﺎب إﻟﻰ ﺑﻌﺾ اﻟﻄﺮق اﻟﺘﻘﻠﻴﺪﻳﺔ اﻟﻬﻮرﺳﺘﻴﻜﻴﺔ ﻟﻠﺘﻨﺒﺆ اﻹﺣﺼﺎﺋﻲ وﻓﻲ ﺟﻤﻴﻊ‬ ‫أﺟﺰاء اﻟﻜﺘﺎب ﻗﻤﺖ ﺑﺘﻮﺿﻴﺢ اﻷﻣﺜﻠﺔ واﻟﺤﺎﻻت اﻟﺪراﺳﻴﺔ ﺑﺈﺳﺘﺨﺪام اﻟﺤﺰﻣﺔ اﻹﺣﺼﺎﺋﻴﺔ‬ ‫‪ Minitab‬وهﻲ ﺑﺮاﻣﺞ ﺣﺎﺳﺐ ﻃﻮرت ﺧﺎﺻﺔ ﻟﺘﻌﻠﻴﻢ ﻋﻠﻢ اﻹﺣﺼﺎء ﺑﺠﻤﻴﻊ ﻓﺮوﻋﻪ‬ ‫وهﺬﻩ اﻟﺤﺰﻣﺔ ﻣﺘﻮﻓﺮة ﻟﻠﻄﻼب ﺑﺎﻟﻤﺠﺎن‪.‬‬ ‫اﻟﺠﺰء اﻟﺜﺎﻧﻲ ﻣﻦ اﻟﻜﺘﺎب وﻣﻮﺟﻪ ﻟﻄﻼب اﻟﺪراﺳﺎت اﻟﻌﻠﻴﺎ ﺳﻮف ﻳﺘﻄﺮق ﺑﺈذن اﷲ‬ ‫ﻟﻤﻮاﺿﻴﻊ ﻣﺜﻞ ﺗﺤﻠﻴﻞ اﻟﺘﺪﺧﻞ ‪ Intervention Analysis‬وﻧﻤﺎذج داﻟﺔ اﻟﺘﺤﻮﻳﻞ‬ ‫‪ Transfer Function Models‬وﻧﻤﺎذج اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﺘﻌﺪدة‬ ‫‪ Multivariate Time Series Models‬وﻧﻤﺎذج اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺟﻬﻪ‬ ‫‪ Vector Time Series Models‬وﻧﻤﺎذج ﻓﻀﺎء اﻟﺤﺎﻟﺔ وﻣﺮﺷﺢ آﺎﻟﻤﻦ ‪State‬‬ ‫‪ Space Models and Kalman Filtering‬وﻧﻤﺎذج اﻟﺤﺪ ‪Threshold Time‬‬ ‫‪ Series Models‬وﻧﻤﺎذج ‪ ARCH‬وﻧﻤﺎذج ‪ GARCH‬وﺗﻄﺒﻴﻘﺎﺗﻬﺎ ﻓﻲ اﻟﺘﻨﺒﺆ‬ ‫اﻟﻤﺎﻟﻲ ‪ Finantial Time Series Forecasting‬آﻤﺎ ﺳﻨﺘﻄﺮق إﻟﻰ اﻟﺸﺒﻜﺎت‬ ‫اﻟﻌﺼﺒﻴﺔ ‪ Neural Networks‬وإﺳﺘﺨﺪاﻣﻬﺎ ﻓﻲ اﻟﺘﻨﺒﺆ اﻹﺣﺼﺎﺋﻲ‪.‬‬ ‫هﺬا وارﺟﻮا ﻣﻦ اﷲ ان ﻳﻮﻓﻘﻨﻲ ﻓﻲ إﻧﺠﺎز هﺬا اﻟﻌﻤﻞ ﻟﻮﺟﻬﻪ اﻟﻜﺮﻳﻢ وﻹﺛﺮاء اﻟﻤﻜﺘﺒﺔ‬ ‫اﻟﻌﺮﺑﻴﺔ اﻟﻔﻘﻴﺮة إﻟﻰ ﻣﺜﻞ هﺬا اﻟﻜﺘﺎب‪.‬‬ ‫ﺳﻴﻜﻮن هﺬا اﻟﻜﺘﺎب ﻣﺠﺎﻧﻲ ﻷي ﻃﺎﻟﺐ ﻋﻠﻢ وهﻮ ﺳﻴﻜﻮن ﻣﺘﻮاﺟﺪ ﻋﻠﻰ ﺷﺒﻜﺔ اﻹﻧﺘﺮﻧﺖ‬ ‫ﻓﻲ اﻟﻤﻮﻗﻊ ‪http://www.abarry.net/or/or٢٢١book١.pdf‬‬ ‫واﷲ اﻟﻤﻮﻓﻖ‪.‬‬ ‫اﻟﻤﺆﻟﻒ‬ ‫د‪ .‬ﻋﺪﻧﺎن ﻣﺎﺟﺪ ﻋﺒﺪ اﻟﺮﺣﻤﻦ ﺑﺮي‬ ‫ﺟﺎﻣﻌﺔ اﻟﻤﻠﻚ ﺳﻌﻮد‬ ‫ذو اﻟﻘﻌﺪة ‪ ١٤٢٢‬هـ‬ ‫ﻳﻨﺎﻳﺮ ‪ ٢٠٠٢‬م‬ ‫‪٣‬‬

‫اﻟﻤﺤﺘﻮﻳﺎت‬ ‫ﻣﻘﺪﻣﺔ‬ ‫‪ -١‬اﻟﻔﺼﻞ اﻷول‪ :‬ﻣﻘﺪﻣﺔ وﺗﻌﺎرﻳﻒ‪١٠..................................................................‬‬ ‫‪ ١-١‬أﻣﺜﻠﺔ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ ‪١٠.......................................................‬‬ ‫‪ ٢-١‬اﻟﻐﺮض ﻣﻦ دراﺳﺔ وﺗﺤﻠﻴﻞ اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ ‪١٠ .................................‬‬ ‫‪ ٣-١‬اﻟﺨﻄﻮات اﻟﻤﺘﺨﺬة ﻟﺒﻨﺎء ﻧﻤﻮذج ﺗﻨﺒﺆ ‪١٠ . ................................................‬‬ ‫‪ ١-٣-١‬ﺗﻌﻴﻴﻦ اﻟﻨﻤﻮذج ‪١٠ .. ...................................................................‬‬ ‫‪ ٢-٣-١‬ﺗﻄﺒﻴﻖ اﻟﻨﻤﻮذج ‪١١ ...................................................................‬‬ ‫‪ ٣-٣-١‬ﺗﺸﺨﻴﺺ وإﺧﺘﺒﺎر اﻟﻨﻤﻮذج ‪١١......................................................‬‬ ‫‪ ٤-٣-١‬ﺗﻮﻟﻴﺪ اﻟﺘﻨﺒﺆات ‪١١.....................................................................‬‬ ‫‪ ٥-٣-١‬إﺳﺘﺨﺪام اﻟﺘﻨﺒﺆات ووﺿﻊ اﻟﻘﺮارات ‪١١ ...........................................‬‬ ‫‪ ٤-١‬ﺗﻌﺎرﻳﻒ وﻣﺒﺎدئ أوﻟﻴﺔ ‪١١.....................................................................‬‬ ‫‪ ١-٤-١‬ﺗﻌﺮﻳﻒ ﻣﺎﺿﻲ أو ﺗﺎرﻳﺦ اﻟﻈﺎهﺮة ‪١١ .............................................‬‬ ‫‪ ٢-٤-١‬ﺗﻌﺮﻳﻒ اﻟﺤﺎﺿﺮ أو اﻵن ‪١١ .........................................................‬‬ ‫‪ ٣-٤-١‬ﺗﻌﺮﻳﻒ أﺧﻄﺎء اﻟﺘﻄﺒﻴﻖ ‪١١ .........................................................‬‬ ‫‪ ٤-٤-١‬ﺗﻌﺮﻳﻒ أﺧﻄﺎء اﻟﺘﻨﺒﺆ ‪١٢ .............................................................‬‬ ‫‪ ٥-٤-١‬ﺗﻌﺮﻳﻒ اﻹﺳﺘﻘﺮار ‪١٢ .................................................................‬‬ ‫‪ ٦-٤-١‬ﺗﻌﺮﻳﻒ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء ‪١٢ .........................................................‬‬ ‫‪ ٧-٤-١‬ﻣﺜﺎل ‪ : ١‬اﻟﻤﺸﻲ اﻟﻌﺸﻮاﺋﻲ ‪١٢ ...................................................‬‬ ‫‪ ٨-٤-١‬ﺗﻌﺮﻳﻒ داﻟﺔ اﻟﺘﻐﺎﻳﺮ اﻟﺬاﺗﻲ ‪١٣ ....................................................‬‬ ‫‪ ٩-٤-١‬ﺗﻌﺮﻳﻒ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ‪١٣ ...................................................‬‬ ‫‪ ١٠-٤-١‬ﻣﺜﺎل ‪ : ٢‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﻀﺠﺔ اﻟﺒﻴﻀﺎء ‪١٣ .........................‬‬ ‫‪ ١١-٤-١‬ﺗﻌﺮﻳﻒ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ‪١٤ ......................................‬‬ ‫‪ ١٢-٤-١‬ﻣﺜﺎل ‪ :٣‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻀﺠﺔ اﻟﺒﻴﻀﺎء ‪١٥..................‬‬ ‫‪ ١٣-٤-١‬ﺗﻌﺮﻳﻒ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﻌﻴﻨﺔ ‪١٦.........................................‬‬ ‫‪ ١٤-٤-١‬ﺗﻌﺮﻳﻒ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻌﻴﻨﺔ ‪١٧ ..............................‬‬ ‫‪ ١٥-٤-١‬ﻣﺜﺎل ‪ :٤‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻌﻴﻨﺔ ‪١٨ .........‬‬ ‫‪ -٢‬اﻟﻔﺼﻞ اﻟﺜﺎﻧﻲ‪ :‬ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ‪ ARMA‬وإﺳﺘﺨﺪاﻣﺎﺗﻬﺎ ﻓﻲ‬ ‫اﻟﺘﻨﺒﺆ ‪٢٢................................................................................................‬‬ ‫‪ ١-٢‬ﺗﻌﺮﻳﻒ ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ )‪٢٢ ........ (p,q‬‬ ‫‪ ٢-٢‬ﺗﻌﺮﻳﻒ ﻋﺎﻣﻞ اﻹزاﺣﺔ اﻟﺨﻠﻔﻲ ‪٢٢.........................................................‬‬ ‫‪ ٣-٢‬ﺗﻌﺮﻳﻒ ﻋﺎﻣﻞ اﻹزاﺣﺔ اﻷﻣﺎﻣﻲ ‪٢٢ .....................................................‬‬ ‫‪ ٤-٢‬ﺗﻌﺮﻳﻒ ﻋﺎﻣﻞ اﻟﺘﻔﺮﻳﻖ ‪٢٢..................................................................‬‬ ‫‪ ٥-٢‬ﺗﻌﺮﻳﻒ ﻋﺎﻣﻞ اﻟﺘﺠﻤﻴﻊ ‪٢٢ .................................................................‬‬ ‫‪ ٦-٢‬أﻣﺜﻠﺔ ‪٢٣ .....................................................................................‬‬ ‫‪ ٧-٢‬ﺧﺼﺎﺋﺺ ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ‪٢٤ .............................‬‬ ‫‪ ١-٧-٢‬ﻧﻤﻮذج )‪٢٤ ........................................................... ARMA(٠،٠‬‬ ‫‪ ٢-٧-٢‬ﻧﻤﻮذج )‪٢٧ .................................................................................. AR(١‬‬ ‫‪ ٣-٧-٢‬ﻧﻤﻮذج )‪٣١ .................................................................... AR(٢‬‬ ‫‪٤‬‬

‫‪ ٤-٧-٢‬ﻧﻤﻮذج )‪٣٦ ..................................................................... MA(١‬‬ ‫‪ ٥-٧-٢‬ﻧﻤﻮذج )‪٣٩ .................................................................... MA(٢‬‬ ‫‪ ٦-٧-٢‬ﻧﻤﻮذج )‪٤٠ ............................................................. ARMA(١،١‬‬ ‫‪ ٧-٧-٢‬ﺧﻮاص ﻧﻤﺎذج )‪٤٧ ................................................ ARMA(p,q‬‬ ‫‪ -٣‬اﻟﻔﺼﻞ اﻟﺜﺎﻟﺚ‪ :‬ﻧﻤﺎذج اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ ﻏﻴﺮ اﻟﻤﺴﺘﻘﺮة ‪٤٩ ......................................‬‬ ‫‪ ١-٣‬ﻋﺪم اﻹﺳﺘﻘﺮار ﻓﻲ اﻟﻤﺘﻮﺳﻂ ‪٤٩ ..............................................................‬‬ ‫‪ ٢-٣‬ﻋﺪم اﻹﺳﺘﻘﺮار ﻓﻲ اﻟﺘﺒﺎﻳﻦ ‪٥٠ ................................................................‬‬ ‫‪ ٣-٣‬ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﺘﻜﺎﻣﻠﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ )‪٥٢ ...... (p,d,q‬‬ ‫‪ ١-٣-٣‬ﻧﻤﻮذج )‪٥٢ ....................................................... ARIMA(١،١،٠‬‬ ‫‪ ٢-٣-٣‬ﻧﻤﻮذج )‪٥٢ ........................................................ ARIMA(٠،١،١‬‬ ‫‪ ٣-٣-٣‬ﻧﻤﻮذج اﻟﻤﺸﻲ اﻟﻌﺸﻮاﺋﻲ ﺑﺈﻧﺠﺮاف ‪٥٣ .................................................‬‬ ‫‪ ٤-٣‬داﻟﺔ اﻷوزان ) ‪ y (B‬وﺗﻤﺜﻴﻞ ﻧﻤﺎذج )‪٥٣ ............................... ARMA(p,q‬‬ ‫‪ ٥-٣‬اﻣﺜﻠﺔ ﻟﺪاﻟﺔ اﻷوزان ﻟﺒﻌﺾ اﻟﻨﻤﺎذج ‪٥٤ .....................................................‬‬ ‫‪ ١-٥-٣‬داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج )‪٥٤ .................................................... AR(١‬‬ ‫‪ ٢-٥-٣‬داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج )‪٥٥ .................................................... MA(١‬‬ ‫‪ ٣-٥-٣‬داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج )‪٥٥ .................................................... AR(٢‬‬ ‫‪ ٤-٥-٣‬داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج )‪٥٦ ................................................... MA(٢‬‬ ‫‪ ٥-٥-٣‬داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج )‪٥٦ ........................................... ARMA(١،١‬‬ ‫‪ ٦-٥-٣‬داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج )‪٥٧ ................................................... ARI(١‬‬ ‫‪ ٧-٥-٣‬داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج اﻟﻤﺸﻲ اﻟﻌﺸﻮاﺋﻲ )‪٥٧ .................. ARIMA(١،٠،١‬‬ ‫‪ ٦-٣‬ﺑﻌﺾ ﺧﻮاص داﻟﺔ اﻷوزان ) ‪٥٨ .................................................... y (B‬‬ ‫‪ -٤‬اﻟﻔﺼﻞ اﻟﺮاﺑﻊ‪ :‬اﻟﺘﻨﺒﺆات ذات ﻣﺘﻮﺳﻂ ﻣﺮﺑﻊ اﻟﺨﻄﺄ اﻷدﻧﻰ ﻟﻨﻤﺎذج )‪٥٩ ....... ARMA(p,q‬‬ ‫‪ ١-٤‬ﻧﻈﺮﻳﺔ ‪ :٢‬أﺧﻄﺎء اﻟﺘﻨﺒﺆ ‪٦١ .........................................................................‬‬ ‫‪ ٢-٤‬ﻣﺠﻤﻮﻋﺔ اﻟﻤﻌﻠﻮﻣﺎت ‪٦٢ ................................................. Information Sets‬‬ ‫‪ ٣-٤‬ﻧﻈﺮﻳﺔ ‪ :٣‬اﻟﻤﺘﻨﺒﺊ ذا ﻣﺘﻮﺳﻂ ﻣﺮﺑﻊ اﻟﺨﻄﺄ اﻷدﻧﻰ ‪٦٢ ............................................‬‬ ‫‪ ٤-٤‬ﻗﺎﻋﺪة ‪٦٢ ............................................................................................. ٢‬‬ ‫‪ ٥-٤‬ﺗﻌﺮﻳﻒ داﻟﺔ اﻟﺘﻨﺒﺆ ‪٦٢ .................................................................................‬‬ ‫‪ ٦-٤‬دوال اﻟﺘﻨﺒﺆ ﻟﻨﻤﺎذج )‪٦٢ ....................................................... ARIMA(p,d,q‬‬ ‫‪ ١-٦-٤‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻨﻤﻮذج )‪٦٣ ............................................................... AR(١‬‬ ‫‪ ٢-٦-٤‬ﺷﺮط اﻹﺳﺘﻤﺮار ‪٦٣ ............................................................................‬‬ ‫‪ ٣-٦-٤‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻨﻤﻮذج )‪٦٤ .............................................................. AR(٢‬‬ ‫‪ ٤-٦-٤‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻨﻤﻮذج )‪٦٥ ................................................. ARIMA(٠،١،١‬‬ ‫‪ ٥-٦-٤‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻨﻤﻮذج )‪٦٦ .............................................................. MA(١‬‬ ‫‪ ٦-٦-٤‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻨﻤﻮذج )‪٦٦ .............................................................. MA(٢‬‬ ‫‪ ٧-٦-٤‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻨﻤﻮذج )‪٦٧ ..................................................... ARMA(١،١‬‬ ‫‪ ٧-٤‬ﺣﺪود اﻟﺘﻨﺒﺆ ‪٦٨ .........................................................................................‬‬ ‫‪ ١-٧-٤‬ﺗﻌﺮﻳﻒ ﻓﺘﺮة ﺗﻨﺒﺆ ﻟﻠﻘﻴﻤﺔ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪٦٨ .....................................................‬‬ ‫‪ ٢-٧-٤‬ﻣﺜﺎل ‪٦٩ ........................................................................................‬‬ ‫‪ -٥‬اﻟﻔﺼﻞ اﻟﺨﺎﻣﺲ‪ :‬ﺗﺼﻤﻴﻢ وﺑﻨﺎء ﻧﻈﺎم ﺗﻨﺒﺆ إﺣﺼﺎﺋﻲ ‪٧١ ............................................‬‬ ‫‪ ١-٥‬ﺗﻌﻴﻴﻦ أو ﺗﺤﺪﻳﺪ اﻟﻨﻤﻮذج ‪٧١ .......................................................................‬‬ ‫‪ ١-١-٥‬ﺗﺜﺒﻴﺖ اﻟﺘﺒﺎﻳﻦ ‪٧١ ..............................................................................‬‬ ‫‪ ٢-١-٥‬إﺧﺘﻴﺎر درﺟﺔ اﻟﺘﻔﺮﻳﻖ ‪٧١ ................................................................ d‬‬

‫‪٥‬‬

‫‪ ٣-١-٥‬ﺗﺤﺪﻳﺪ ‪٧١ ............................................................................... p,q‬‬ ‫‪ ٤-١-٥‬إﺿﺎﻓﺔ ﻣﻌﻠﻢ إﻧﺠﺮاف ‪٧١ ...................................................................‬‬ ‫‪ ٢-٥‬ﺗﻘﺪﻳﺮ اﻟﻨﻤﻮذج ‪٧٢ ..............................................................................‬‬ ‫‪ ١-٢-٥‬ﻃﺮﻳﻘﺔ اﻟﻌﺰوم ‪٧٢ ........................................................................‬‬ ‫‪ ٢-٢-٥‬ﺗﻘﺪﻳﺮ اﻟﻌﺰوم ﻟﺒﻌﺾ اﻟﻨﻤﺎذج ‪٧٣ .......................................................‬‬ ‫‪ ١-٢-٢-٥‬ﻟﻨﻤﻮذج )‪٧٣ .............................................................. AR(١‬‬ ‫‪ ٢-٢-٢-٥‬ﻟﻨﻤﻮذج )‪٧٤ ......................................................... MA(١‬‬ ‫‪ ٣-٢-٢-٥‬ﻟﻨﻤﻮذج )‪٧٤ .......................................................... AR(٢‬‬ ‫‪ ٤-٢-٢-٥‬ﻟﻨﻤﻮذج )‪٧٤ ......................................................... MA(٢‬‬ ‫‪ ٥-٢-٢-٥‬ﻟﻨﻤﻮذج )‪٧٥ ................................................ ARMA(١،١‬‬ ‫‪ ٣-٢-٥‬ﻃﺮﻳﻘﺔ اﻟﻤﺮﺑﻌﺎت اﻟﺪﻧﻴﺎ اﻟﺸﺮﻃﻴﺔ ‪٧٥ ..............................................‬‬ ‫‪ ٤-٢-٥‬ﺗﻘﺪﻳﺮات اﻟﻤﺮﺑﻌﺎت اﻟﺪﻧﻴﺎ اﻟﺸﺮﻃﻴﺔ ﻟﺒﻌﺾ اﻟﻨﻤﺎذج ‪٧٦ .........................‬‬ ‫‪ ١-٤-٢-٥‬ﻟﻨﻤﺎذج )‪٧٦ .........................................................AR(١‬‬ ‫‪ ٢-٤-٢-٥‬ﻟﻨﻤﺎذج )‪٧٧ ....................................................... MA(١‬‬ ‫ﺗﺸﺨﻴﺺ وإﺧﺘﺒﺎر اﻟﻨﻤﻮذج‪٧٨ .........................................................‬‬ ‫‪٣-٥‬‬ ‫‪ ١-٣-٥‬ﻓﺤﺺ اﻟﺒﻮاﻗﻲ ‪٧٨ ....................................................................‬‬ ‫‪ ١-١-٣-٥‬إﺧﺘﺒﺎر اﻟﻤﺘﻮﺳﻂ ﻟﻠﺒﻮاﻗﻲ ‪٧٩ ..................................................‬‬ ‫‪ ٢-١-٣-٥‬إﺧﺘﺒﺎر اﻟﻌﺸﻮاﺋﻴﺔ ﻟﻠﺒﻮاﻗﻲ ‪٧٩ .................................................‬‬ ‫‪ ٣-١-٣-٥‬إﺧﺘﺒﺎر اﻟﺘﺮاﺑﻂ أو اﻹﺳﺘﻘﻼل ﻟﻠﺒﻮاﻗﻲ ‪٧٩ ...................................‬‬ ‫‪ ٤-١-٣-٥‬إﺧﺘﺒﺎر ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ ‪٧٩ ....................................................‬‬ ‫‪ ٢-٣-٥‬ﺑﻌﺾ اﻟﻤﻌﺎﻳﻴﺮ اﻻﺧﺮى ﻹﺧﺘﻴﺎر ﻧﻤﻮذج ﻣﻨﺎﺳﺐ ‪٧٩ .............................‬‬ ‫‪ ١-٢-٣-٥‬إﺣﺼﺎﺋﻴﺔ آﻴﻮ ﻟﻠﺠﻨﻖ وﺑﻜﺲ ‪٧٩ ...........................................‬‬ ‫‪ ٢-٢-٣-٥‬ﻣﻌﻴﺎر اﻹﻋﻼم اﻟﺬاﺗﻲ ‪٨٠ ......................................... AIC‬‬ ‫‪ ٣-٣-٥‬أﻣﺜﻠﺔ وﺣﺎﻻت دراﺳﺔ ‪٨٠ ..........................................................‬‬ ‫‪ -٦‬اﻟﻔﺼﻞ اﻟﺴﺎدس‪ :‬ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﺘﻜﺎﻣﻠﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك اﻟﻤﻮﺳﻤﻴﺔ ‪١١٣ ......‬‬ ‫‪ ١-٦‬دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﺒﻌﺾ اﻟﻨﻤﺎذج اﻟﻤﻮﺳﻤﻴﺔ ‪١١٤ .....‬‬ ‫‪ ١-١-٦‬ﻟﻨﻤﻮذج ‪١١٤ ......................................... SARMA(٠،١)(١،١)١٢‬‬ ‫‪ ٢-٦‬دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﺒﻌﺾ اﻟﻨﻤﺎذج اﻟﻤﻮﺳﻤﻴﺔ ‪١١٥ ..................................‬‬ ‫‪١١٥............................................. SARIMA(٠,d,٠)(٠,D,١)s ١-٢-٦‬‬ ‫‪١١٥ ........................................... SARIMA(٠,d,٠)(١,D,١)s ٢-٢-٦‬‬ ‫‪١١٥ ............................................ SARIMA(٠,d,١)(٠,D,١)s ٣-٢-٦‬‬ ‫‪١١٦ ........................................... SARIMA(٠,d,٠)(١,D,١)s ٤-٢-٦‬‬ ‫‪١١٦ ........................................... SARIMA(٠,d,١)(١,D,٠)s ٥-٢-٦‬‬ ‫‪١١٦ ............................................ SARIMA(٠,d,٢)(٠,D,١)s ٦-٢-٦‬‬ ‫‪ ٣-٦‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻨﻤﻮذج اﻟﻤﻮﺳﻤﻲ اﻟﺘﻀﺎﻋﻔﻲ ‪١١٨ ....................‬‬ ‫‪ ٤-٦‬أﻣﺜﻠﺔ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ ‪١١٩ .......................................‬‬ ‫‪ ٥-٦‬إﺷﺘﻘﺎق دوال ﺗﻨﺒﺆ ﻟﺒﻌﺾ ﻧﻤﺎذج اﻟﻤﺘﺴﻠﺴﻼت اﻟﻤﻮﺳﻤﻴﺔ اﻟﺘﻀﺎﻋﻔﻴﺔ ‪١٢٣ ..........‬‬ ‫‪ ١-٥-٦‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻠﻨﻤﻮذج ‪١٢٣ .................... SARIMA(٠،٠،٠)(٠،١،١)١٢‬‬ ‫‪ ٢-٥-٦‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻠﻨﻤﻮذج ‪١٢٤ .................... SARIMA(٠،١،١)(٠،١،١)١٢‬‬ ‫‪ ٦-٦‬أﻣﺜﻠﺔ وﺣﺎﻻت دراﺳﺔ ﻟﻠﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ ‪١٢٥ ..........................‬‬ ‫اﻟﺠﺰء اﻟﻌﻤﻠﻲ‪:‬‬ ‫‪ -٧‬اﻟﻔﺼﻞ اﻟﺴﺎﺑﻊ‪ :‬ورﻗﺔ ﺗﺪرﻳﺐ ﻋﻤﻠﻲ ﻋﻠﻰ اﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﻤﺘﻮﺳﻂ‬ ‫اﻟﻤﺘﺤﺮك ‪١٤٢ ............................................................................................‬‬ ‫‪٦‬‬

‫‪ -٨‬اﻟﻔﺼﻞ اﻟﺜﺎﻣﻦ‪ :‬ﻣﺜﺎل ﺗﺤﻠﻴﻞ اﻟﺒﻮاﻗﻲ وﻣﻌﻴﻴﺮ إﺧﺘﻴﺎر ﻧﻤﻮذج ﻣﻨﺎﺳﺐ ‪١٥٥ ..................‬‬ ‫‪ -٩‬اﻟﻔﺼﻞ اﻟﺘﺎﺳﻊ‪ :‬ﺗﺤﻠﻴﻞ أو ﺗﻔﻜﻴﻚ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ إﻟﻰ ﻣﺮآﺒﺎت ‪١٦٢ .....................‬‬ ‫‪ -١٠‬اﻟﺘﻤﻬﻴﺪ واﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ‪١٧٨ ...........................................‬‬ ‫‪ ١-١٠‬اﻟﻮﺳﻴﻂ اﻟﺠﺎري ‪١٨١ .............................................................‬‬ ‫‪ -١١‬اﻟﻔﺼﻞ اﻟﺤﺎدي ﻋﺸﺮ‪ :‬اﻟﺘﻤﻬﻴﺪ واﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ اﻟﺘﻤﻬﻴﺪ اﻷﺳﻲ اﻟﺒﺴﻴﻂ ‪١٨٤ .............‬‬ ‫‪ -١٢‬اﻟﻔﺼﻞ اﻟﺜﺎﻧﻲ ﻋﺸﺮ‪ :‬اﻟﺘﻤﻬﻴﺪ واﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ اﻟﺘﻤﻬﻴﺪ اﻷﺳﻲ اﻟﻤﺰدوج ‪١٩٠ ...........‬‬ ‫‪ ١-١٢‬ﻃﺮﻳﻘﺔ ﺑﺮاون ‪١٩٠ ........................................................................‬‬ ‫‪ ٢-١٢‬ﻃﺮﻳﻘﺔ هﻮﻟﺖ ‪١٩٠ ........................................................................‬‬ ‫‪ ٣-١٢‬أﻣﺜﻠﺔ ‪١٩١ ..................................................................................‬‬ ‫‪ -١٣‬اﻟﻔﺼﻞ اﻟﺜﺎﻟﺚ ﻋﺸﺮ‪ :‬اﻟﺘﻤﻬﻴﺪ اﻷﺳﻲ اﻟﺜﻼﺛﻲ واﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ ﻃﺮﻳﻘﺔ وﻧﺘﺮز ﻟﻠﻤﺘﺴﻠﺴﻼت‬ ‫اﻟﻤﻮﺳﻤﻴﺔ اﻟﻤﻨﺠﺮﻓﺔ ‪١٩٨ .............................................................................‬‬ ‫‪ ١-١٣‬اﻟﻨﻤﻮذج اﻹﺿﺎﻓﻲ ‪١٩٩ ...............................................................‬‬ ‫‪ ٢-١٣‬اﻟﻨﻤﻮذج اﻟﺘﻀﺎﻋﻔﻲ ‪٢٠١ ..............................................................‬‬ ‫‪ ٣-١٣‬ﻣﺜﺎل ﻟﺒﻨﺎء ﻧﻤﻮذج ﺗﻨﺒﺆ ‪٢٠٤ ...........................................................‬‬ ‫‪ ٤-١٣‬ﻣﺜﺎل ﺁﺧﺮ ﻟﺒﻨﺎء ﻧﻤﻮذج ﺗﻨﺒﺆ ‪٢١٠ .....................................................‬‬ ‫اﻟﻤﺮاﺟﻊ ‪٢١٩ ...........................................................................................‬‬

‫‪٧‬‬

‫اﻟﻔﺼﻞ اﻷول‬ ‫ﻣﻘﺪﻣﺔ وﺗﻌﺎرﻳﻒ‬ ‫ﺗﻌﺮﻳﻒ ‪ :١‬اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ ‪Time Series‬‬

‫ﻣﺘﺘﺎﺑﻌﺔ ﻣﻦ اﻟﻘﻴﻢ اﻟﻤﺸﺎهﺪة ﻟﻈﺎهﺮة ﻋﺸﻮاﺋﻴﺔ ﻣﺮﺗﺒﺔ ﻣﻊ اﻟﺰﻣﻦ ) او ﻣﺮﺗﺒﺔ ﻣﻊ اﻟﻤﻜﺎن (‬

‫اﻣﺜﻠﺔ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ‪:‬‬ ‫‪ -١‬ﺳﻌﺮ اﻗﻔﺎل ﺳﻬﻢ ﺑﻨﻚ اﻟﺮﻳﺎض ﻳﻮﻣﻴﺎ‪.‬‬ ‫‪ -٢‬ﻋﺪد اﻟﻮﺣﺪات اﻟﻤﻄﻠﻮﺑﺔ اﺳﺒﻮﻋﻴﺎ ﻣﻦ اﻧﺘﺎج ﺳﻠﻌﺔ ﻣﻌﻴﻨﺔ‪.‬‬ ‫‪ -٣‬ﺣﺠﻢ اﻟﻤﺒﻴﻌﺎت ﺷﻬﺮﻳﺎ ﻣﻦ ﺳﻠﻌﺔ ﻣﺎ‪.‬‬ ‫‪ -٤‬ﺣﺠﻢ اﻹﻧﺘﺎج اﻟﻴﻮﻣﻲ ﻟﻠﻨﻔﻂ اﻟﺨﺎم ﺑﺎﻟﻤﻤﻠﻜﺔ‪.‬‬ ‫واﻟﻐﺮض ﻣﻦ دراﺳﺔ وﺗﺤﻠﻴﻞ اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ هﻮ‪:‬‬ ‫‪ -١‬ﻓﻬﻢ وﻧﻤﺬﺟﺔ ﻋﺸﻮاﺋﻴﺔ اﻟﻈﺎهﺮة اﻟﻤﺸﺎهﺪة‪.‬‬ ‫‪ -٢‬اﻟﺘﻨﺒﺆ ﻋﻦ اﻟﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﻟﻠﻈﺎهﺮة اﻟﻌﺸﻮاﺋﻴﺔ‪.‬‬ ‫‪ -٣‬اﻟﺘﺤﻜﻢ ﺑﺎﻟﻈﺎهﺮة اﻟﻌﺸﻮاﺋﻴﺔ إذا اﻣﻜﻦ ذﻟﻚ‪.‬‬ ‫اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ ﻟﻤﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺸﺎهﺪة وهﻲ ﻋﺒﺎرة ﻋﻦ اﻹﻧﺘﺎج اﻟﻴﻮﻣﻲ ﻟﻠﺤﻠﻴﺐ ﺑﺎﻟﺮﻃﻞ ﻟﺒﻘﺮة ﻣﺎ‬

‫‪850‬‬

‫‪750‬‬

‫‪C1‬‬ ‫‪650‬‬

‫‪550‬‬

‫‪70‬‬

‫‪60‬‬

‫‪50‬‬

‫‪30‬‬

‫‪40‬‬

‫‪20‬‬

‫‪10‬‬

‫‪Index‬‬

‫اﻟﺨﻄﻮات اﻟﻤﺘﺨﺬة ﻟﺒﻨﺎء ﻧﻤﻮذج ﺗﻨﺒﺆ‪:‬‬ ‫إن إﻳﺠﺎد ﻧﻤﻮذج ﻣﻨﺎﺳﺐ ﺗﻨﻄﺒﻖ ﻋﻠﻴﺔ ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺸﺎهﺪة ﻳﻌﺘﺒﺮ ﻣﻦ اﻟﻤﻬﺎم اﻟﺼﻌﺒﺔ واﻟﺘﻲ ﺗﺤﺘﺎج‬ ‫اﻟﻰ اﻟﻜﺜﻴﺮ ﻣﻦ اﻟﺒﺤﺚ واﻟﺨﺒﺮة‪ .‬ﺳﻮف ﻧﺴﺘﻌﺮض ﺑﻌﺾ اﻟﺨﻄﻮات اﻟﻌﺮﻳﻀﺔ ﻟﺒﻨﺎء ﻧﻤﻮذج رﻳﺎﺿﻲ‬ ‫ﻟﻠﺘﻨﺒﺆ ﻋﻦ ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺎ‪:‬‬ ‫‪-١‬‬ ‫ﺗﻌﻴﻴﻦ اﻟﻨﻤﻮذج أو ﺗﺤﺪﻳﺪ اﻟﻨﻤﻮذج ‪ :Model Identification‬وهﺬا ﻳﺘﻢ ﺑﺮﺳﻢ‬ ‫اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ ﻓﻴﻤﺎ ﻳﺴﻤﻰ ‪ Time Plot‬ﺣﻴﺚ ﻳﻜﻮن اﻹﺣﺪاﺛﻲ اﻻﻓﻘﻲ هﻮ اﻟﺰﻣﻦ واﻟﺮأﺳﻲ ﺣﺠﻢ‬ ‫اﻟﻈﺎهﺮة اﻟﻤﺸﺎهﺪة وﻣﻦ ﺛﻢ إﺧﺘﻴﺎر ﻧﻤﻮذج رﻳﺎﺿﻲ ﻣﻌﺘﻤﺪﻳﻦ ﻋﻠﻲ ﺑﻌﺾ اﻟﻤﻘﺎﻳﻴﺲ اﻹﺣﺼﺎﺋﻴﺔ اﻟﺘﻰ‬ ‫ﺗﻤﻴﺰ ﻧﻤﻮذج ﻋﻦ ﺁﺧﺮ وﻋﻠﻰ اﻟﺨﺒﺮة اﻟﻤﺴﺘﻤﺪة ﻣﻦ اﻟﺪراﺳﺎت واﻻﺑﺤﺎث‪.‬‬

‫‪٨‬‬

‫‪-٢‬‬

‫‪-٣‬‬

‫‪-٤‬‬

‫‪-٥‬‬

‫ﺗﻄﺒﻴﻖ اﻟﻨﻤﻮذج ‪ :Model Fitting‬ﺑﻌﺪ ﺗﺮﺷﻴﺢ ﻧﻤﻮذج او اآﺜﺮ آﻨﻤﻮذج ﻣﻨﺎﺳﺐ ﻟﻮﺻﻒ‬ ‫اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﺸﺎهﺪة ﻧﻘﻮم ﺑﺘﻘﺪﻳﺮ ﻣﻌﺎﻟﻢ هﺬا اﻟﻨﻤﻮذج ﻣﻦ اﻟﺒﻴﺎﻧﺎت اﻟﻤﺸﺎهﺪة ﺑﺈﺳﺘﺨﺪام ﻃﺮق‬ ‫اﻟﺘﻘﺪﻳﺮ اﻹﺣﺼﺎﺋﻲ اﻟﺨﺎﺻﺔ ﺑﺎﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ وهﺬا اﻟﻨﻤﻮذج اﻟﻤﺮﺷﺢ ﻳﺆﺧﺬ آﻨﻤﻮذج‬ ‫اوﻟﻲ ﻗﺎﺑﻞ ﻟﻠﺘﻌﺪﻳﻞ ﻻﺣﻘﺎ‪.‬‬ ‫ﺗﺸﺨﻴﺺ وإﺧﺘﺒﺎر اﻟﻨﻤﻮذج ‪ :Model Diagnostics‬إﺟﺮاء إﺧﺘﺒﺎرات ﺗﻔﺤﺼﻴﺔ ﻋﻠﻰ‬ ‫أﺧﻄﺎء اﻟﺘﻄﺒﻴﻖ ‪ Fitting Errors‬ﻟﻤﻌﺮﻓﺔ ﻣﺪى ﺗﻄﺎﺑﻖ اﻟﻤﺸﺎهﺪات ﻣﻊ اﻟﻘﻴﻢ اﻟﻤﺤﺴﻮﺑﺔ ﻣﻦ‬ ‫اﻟﻨﻤﻮذج اﻟﻤﺮﺷﺢ وﻣﺪى ﺻﺤﺔ ﻓﺮﺿﻴﺎت اﻟﻨﻤﻮذج‪ .‬ﻓﻲ ﺣﺎﻟﺔ إﺟﺘﻴﺎز اﻟﻨﻤﻮذج اﻟﻤﺮﺷﺢ ﻟﻬﺬﻩ‬ ‫اﻹﺧﺘﺒﺎرات ﻧﻘﻮم ﺑﺈﻋﺘﻤﺎدة ﻋﻠﻰ اﻧﻪ اﻟﻨﻤﻮذج اﻟﻨﻬﺎﺋﻲ وﻳﺴﺘﺨﺪم ﻟﺘﻮﻟﻴﺪ ﺗﻨﺒﺆات ﻟﻠﻘﻴﻢ‬ ‫اﻟﻤﺴﺘﻘﺒﻠﻴﺔ وإﻻ ﻧﻌﻮد ﻟﻠﺨﻄﻮة اﻻوﻟﻰ ﻟﺘﻌﻴﻴﻦ ﻧﻤﻮذج ﺟﺪﻳﺪ‪.‬‬ ‫ﺗﻮﻟﻴﺪ اﻟﺘﻨﺒﺆات ‪ :Forecast Generation‬ﻳﺴﺘﺨﺪم اﻟﻨﻤﻮذج اﻟﻨﻬﺎﺋﻲ ﻟﺘﻮﻟﻴﺪ ﺗﻨﺒﺆات ﻋﻦ‬ ‫اﻟﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ وﻣﻦ ﺛﻢ ﺣﺴﺎب أﺧﻄﺎء اﻟﺘﻨﺒﺆ ‪ Forecast Errors‬آﻠﻤﺎ اﺳﺘﺠﺪت ﻗﻴﻢ‬ ‫ﺟﺪﻳﺪة ﻣﺸﺎهﺪة ﻣﻦ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ وﻣﺮاﻗﺒﺔ هﺬﻩ اﻷﺧﻄﺎء ﻓﻰ ﻣﺎ ﻳﺴﻤﻰ ﺑﻤﺨﻄﻄﺎت‬ ‫اﻟﻤﺮاﻗﺒﺔ ‪ Control Charts‬واﻟﺘﻲ ﺗﻮﺿﻊ ﻟﻠﻘﺒﻮل ﺑﻨﺴﺒﺔ ﺧﻄﺄ ﻣﻌﻴﻦ إذا ﺗﺠﺎوزﺗﺔ أﺧﻄﺎء‬ ‫اﻟﺘﻨﺒﺆ ﻳﻌﺎد اﻟﻨﻈﺮ ﻓﻲ اﻟﻨﻤﻮذج وﺗﻌﺎد اﻟﺪورة ﻣﻦ ﺟﺪﻳﺪ ﺑﺘﺤﺪﻳﺪ ﻧﻤﻮذج ﻣﺮﺷﺢ ﺁﺧﺮ‪.‬‬ ‫إﺳﺘﺨﺪام اﻟﺘﻨﺒﺆات ووﺿﻊ اﻟﻘﺮارات ‪:Implementation and Decision making‬‬ ‫ﺗﻘﺪم اﻟﺘﻨﺒﺆات ﻓﻰ ﺗﻘﺮﻳﺮ ﻟﺼﺎﻧﻌﻲ اﻟﻘﺮار ﻟﻠﻨﻈﺮ ﻓﻲ إﺳﺘﺨﺪاﻣﻬﺎ ﺑﺎﻟﺸﻜﻞ اﻟﻤﻨﺎﺳﺐ‪.‬‬

‫ﺗﻌﺎرﻳﻒ وﻣﺒﺎدئ اوﻟﻴﺔ‪:‬‬ ‫ﺳﻮف ﻧﺮﻣﺰ ﻟﻠﻈﺎهﺮة اﻟﻌﺸﻮاﺋﻴﺔ أو اﻟﻌﻤﻠﻴﺔ اﻟﻌﺸﻮاﺋﻴﺔ اﻟﺘﻲ ﺗﻮﻟﺪ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ ﺑﺎﻟﺮﻣﺰ‬ ‫}‪ {L, Z −1 , Z 0 , Z1 , Z 2 ,L‬او اﺧﺘﺼﺎرا }}‪ {Z t , t ∈{L, −1, 0,1, 2,L‬او ﺑﺒﺴﺎﻃﺔ‬

‫وﻟﻠﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ اﻟﻤﺸﺎهﺪة ﺑﺎﻟﺮﻣﺰ } ‪{z1 , z2 ,L, zn−1 , zn‬‬

‫} ‪{Z t‬‬

‫ﺗﻌﺮﻳﻒ‪ :٢‬اﻟﻘﻴﻢ ‪ z1 , z2 ,L , zn −1‬ﺗﺴﻤﻰ ﺑﺎﻟﻤﺎﺿﻰ او ﺗﺎرﻳﺦ اﻟﻈﺎهﺮة ‪History‬‬

‫واﻟﺘﺎرﻳﺦ ﻣﻬﻢ ﺟﺪا ﻓﻲ ﻋﻤﻠﻴﺔ اﻟﻨﻤﺬﺟﺔ‬

‫ﺗﻌﺮﻳﻒ ‪ : ٣‬اﻟﻘﻴﻤﺔ ‪ zn‬ﺗﺴﻤﻰ اﻟﺤﺎﺿﺮ او اﻵن‬ ‫وهﻲ اﻟﻤﺸﺎهﺪة اﻷﺧﻴﺮة ‪.‬‬

‫ﺗﻌﺮﻳﻒ ‪ :٤‬أﺧﻄﺎء اﻟﺘﻄﺒﻴﻖ ﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ ‪ et = zt − zˆt , t = 1, 2,..., n‬ﺣﻴﺚ ‪ zˆt‬هﻲ اﻟﻘﻴﻢ‬ ‫اﻟﻤﻄﺒﻘﺔ ) اﻟﻘﻴﻢ اﻟﺘﻲ ﻧﺘﺤﺼﻞ ﻋﻠﻴﻬﺎ ﻣﻦ اﻟﻨﻤﻮذج( وﺗﺴﻤﻲ أﻳﻀﺎ اﻟﺮواﺳﺐ ‪Residuals‬‬ ‫وﻳﻼﺣﻆ ان اﺧﻄﺎء اﻟﺘﻄﺒﻴﻖ ﻧﺤﺼﻞ ﻋﻠﻴﻬﺎ دﻓﻌﺔ واﺣﺪة ﺑﻌﺪ ﺗﻘﺪﻳﺮ اﻟﻨﻤﻮذج‪.‬‬ ‫ﻣﻼﺣﻈﺔ‪ :‬ﺳﻮف ﻧﺮﻣﺰ ﻟﻠﻤﺸﺎهﺪات اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﺑﺎﻟﺮﻣﻮز ‪ zn +1 , zn + 2 , zn +3 ,...‬او ﺑﺸﻜﻞ ﻋﺎم‬ ‫‪ zn +l , l ≥ 0‬وﻧﺮﻣﺰ ﻟﺘﻨﺒﺆاﺗﻬﺎ ﺑﺎﻟﺮﻣﺰ ‪ zn (1) , zn ( 2 ) , zn ( 3) ,...‬او ﺑﺸﻜﻞ ﻋﺎم‬ ‫‪zn ( l ) , l ≥ 0‬‬

‫‪٩‬‬

‫‪en ( l ) = zn +l − zn ( l ) , l ≥ 0‬‬

‫ﺗﻌﺮﻳﻒ ‪ :٥‬أﺧﻄﺎء اﻟﺘﻨﺒﺆ ﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬

‫وأﺧﻄﺎء اﻟﺘﻨﺒﺆ ﺗﻨﺘﺞ اﻟﻮاﺣﺪة ﺗﻠﻮ اﻻﺧﺮى آﻠﻤﺎ ﺗﻘﺪم اﻟﺰﻣﻦ وﺷﻮهﺪت اﻟﻘﻴﻢ اﻟﺤﻘﻴﻘﻴﺔ‬ ‫ﺗﻌﺮﻳﻒ ‪ :٦‬ﻳﻘﺎل ان اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ اﻟﻤﺸﺎهﺪة } ‪ {z1 , z2 ,L, zn−1 , zn‬ﻣﺴﺘﻘﺮة ‪Stationary‬‬

‫إذا ﺣﻘﻘﺖ اﻟﺸﺮوط اﻟﺘﺎﻟﻴﺔ‪:‬‬

‫‪1) E ( zt ) = constant = µ , ∀t‬‬ ‫‪⎧constant = γ 0 , ∀t , ∀s, t = s‬‬ ‫⎨ = ) ‪2) cov ( zt , zs‬‬ ‫‪⎩ f ( s − t ) , ∀t , ∀ s , t ≠ s‬‬

‫اﻵن ﺳﻮف ﻧﻌﺮف ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﻬﻤﺔ ﺟﺪا ﻟﻜﻮﻧﻬﺎ ﺣﺠﺮة او ﻃﻮب اﻟﺒﻨﺎء ‪Building Blocks‬‬ ‫ﻟﺠﻤﻴﻊ اﻟﻨﻤﺎذج اﻟﺘﻲ ﺳﻮف ﻧﺪرﺳﻬﺎ‬

‫ﺗﻌﺮﻳﻒ ‪ :٧‬ﻣﺘﺴﻠﺴﻠﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء ‪ White Noise Series‬اوﻋﻤﻠﻴﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‬ ‫‪ {at } White Noise Process‬هﻲ ﻋﺒﺎرة ﻋﻦ ﻣﺘﺘﺎﺑﻌﺔ ﻣﻦ اﻟﻤﺸﺎهﺪات اﻟﻌﺸﻮاﺋﻴﺔ ﻏﻴﺮ‬ ‫اﻟﻤﺘﺮاﺑﻄﺔ ) واﺣﻴﺎﻧﺎ ﻧﻔﺘﺮض اﻧﻬﺎ ﻣﺘﺘﺎﺑﻌﺔ ﻣﻦ اﻟﻤﺘﻐﻴﺮات اﻟﻌﺸﻮاﺋﻴﺔ اﻟﺘﻲ ﺗﻜﻮن ﻣﺴﺘﻘﻠﺔ وﻟﻬﺎ‬ ‫ﺗﻮزﻳﻌﺎت ﻣﺘﻄﺎﺑﻘﺔ )‪ ( Independent, Identically Distributed (IID‬ﺑﻤﺘﻮﺳﻂ ﺻﻔﺮي‬ ‫وﺗﺒﺎﻳﻦ ﺛﺎﺑﺖ ‪ σ 2‬أي‪:‬‬

‫‪1) E ( at ) = 0, ∀t‬‬

‫وﻳﺮﻣﺰ ﻟﻬﺎ ﺑﺎﻟﺮﻣﺰ ) ‪WN ( 0, σ 2‬‬

‫‪⎧σ 2 , ∀t , ∀s, t = s‬‬ ‫⎨ = ) ‪2) cov ( at , as‬‬ ‫‪⎩ 0 , ∀t , ∀ s , t ≠ s‬‬ ‫‪at‬‬

‫ﻣﺜﺎل‪ :١‬ﻣﺘﺴﻠﺴﻠﺔ اﻟﻤﺸﻲ اﻟﻌﺸﻮاﺋﻲ ‪: Random Walk‬‬ ‫ﺳﻮف ﻧﺒﻨﻲ ﻋﻤﻠﻴﺔ ﻋﺸﻮاﺋﻴﺔ } ‪ {Z t‬آﺎﻟﺘﺎﻟﻲ‪:‬‬

‫‪Z1 = a1‬‬ ‫‪Z 2 = a1 + a2‬‬ ‫‪M‬‬

‫‪Z t = a1 + a2 +L + at‬‬ ‫أو‬

‫‪Z t = Z t −1 + at‬‬ ‫أي ﻟﻮ اﻋﺘﺒﺮﻧﺎ ان ‪ a j‬هﻮ ﺣﺠﻢ اﻟﺨﻄﻮة اﻟﺘﻲ ﺗﺆﺧﺬ اﻟﻲ اﻻﻣﺎم او اﻟﺨﻠﻒ ﻋﻨﺪ اﻟﺰﻣﻦ ‪ j‬ﻓﺎن ‪Z t‬‬

‫هﻲ ﻣﻮﻗﻊ ﻣﺎﺷﻲ ﻋﺸﻮاﺋﻲ ﻋﻨﺪ اﻟﺰﻣﻦ ‪t‬‬ ‫ﻣﻼﺣﻈﺔ‪ :‬هﺬﻩ اﻟﻌﻤﻠﻴﺔ او اﻟﻤﺘﺴﻠﺴﻠﺔ ﻣﻦ اﻟﻨﻤﺎذج اﻟﻬﺎﻣﺔ ﺟﺪا اﻟﺘﻲ ﺗﺼﻒ اﺳﻮاق اﻟﻤﺎل اﻟﻌﺎﻟﻤﻴﺔ‬ ‫ﺗﻤﺮﻳﻦ‪ :‬اوﺟﺪ ) ‪ E ( Z t‬و ) ‪ cov ( Z t , Z s‬ﻟﺠﻤﻴﻊ ﻗﻴﻢ ‪ t , s‬وهﻞ اﻟﻌﻤﻠﻴﺔ ﻣﺴﺘﻘﺮة؟‬ ‫‪١٠‬‬

‫ﺗﻌﺮﻳﻒ ‪ :٨‬داﻟﺔ اﻟﺘﻐﺎﻳﺮ اﻟﺬاﺗﻲ ‪ Autocovariance Function‬وﺗﻌﺮف آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫‪= cov ( Z t , Z s ) , ∀t , ∀s‬‬

‫‪γ t ,s‬‬

‫‪= E ⎡⎣( Z t − µ )( Z s − µ ) ⎤⎦ , ∀t , ∀s‬‬

‫وإذا ﻋﺮﻓﻨﺎ اﻟﺘﺨﻠﻒ ‪ k‬ﻋﻠﻲ اﻧﻪ اﻟﻔﺘﺮة اﻟﺰﻣﻨﻴﺔ اﻟﺘﻲ ﺗﻔﺼﻞ ﺑﻴﻦ ‪ Z t‬وﺑﻴﻦ ‪ Z t −k‬أو ‪ Z t +k‬ﻓﺈن‬ ‫داﻟﺔ اﻟﺘﻐﺎﻳﺮ اﻟﺬاﺗﻲ ﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‪:‬‬ ‫‪γ k = cov ( Z t , Z t −k ) , k = 0, ±1, ±2,L‬‬ ‫‪= E ⎡⎣( Z t − µ )( Z t −k − µ ) ⎤⎦ , k = 0, ±1, ±2,L‬‬ ‫ﻣﻼﺣﻈﺔ‪ :‬ﺳﻮف ﻧﺴﺘﺨﺪم اﻟﺘﻌﺮﻳﻒ اﻟﺜﺎﻧﻲ داﺋﻤﺎ‬

‫ﺗﻌﺮﻳﻒ ‪ :٩‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ )‪ Autocorrelation Function (ACF‬وﺗﻌﺮف آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫‪γ‬‬ ‫‪ρ k = k , k = 0, ±1, ±2,L‬‬ ‫‪γ0‬‬ ‫وﻟﻬﺎ اﻟﺨﻮاص اﻟﺘﺎﻟﻴﺔ‪:‬‬

‫‪1. ρ 0 = 1‬‬ ‫‪2. ρ − k = ρ k‬‬

‫‪ρk ≤ 1‬‬

‫‪3.‬‬

‫ﻣﺜﺎل ‪ :٢‬ﺳﻮف ﻧﺸﺘﻖ اﻵن داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻌﻤﻠﻴﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‬ ‫داﻟﺔ اﻟﺘﻐﺎﻳﺮ اﻟﺬاﺗﻲ ﻟﻌﻤﻠﻴﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء هﻲ‪:‬‬ ‫‪2‬‬ ‫‪⎧σ , k = 0‬‬ ‫⎨ = ) ‪γ k = cov ( at , at −k‬‬ ‫‪⎩ 0, k ≠0‬‬ ‫وذﻟﻚ ﻣﻦ اﻟﺘﻌﺮﻳﻒ ‪ ٧‬وﻣﻨﻬﺎ ﻧﺠﺪ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ‪:‬‬ ‫‪⎧1 , k = 0‬‬ ‫‪γ‬‬ ‫⎨ = ‪ρk = k‬‬ ‫‪γ 0 ⎩0 , k ≠ 0‬‬ ‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‪:‬‬ ‫‪Autocorrelation function of White Noise‬‬ ‫‪1.0‬‬

‫‪0.0‬‬ ‫‪9‬‬

‫‪8‬‬

‫‪7‬‬

‫‪6‬‬

‫‪5‬‬

‫‪4‬‬

‫‪Lag‬‬

‫‪١١‬‬

‫‪3‬‬

‫‪2‬‬

‫‪1‬‬

‫‪0‬‬

‫‪Autocorr‬‬

‫‪0.5‬‬

‫ﺗﻌﺮﻳﻒ ‪ :١٠‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ‪Partial Autocorrelation Function‬‬ ‫)‪(PACF‬‬ ‫وﺗﻌﻄﻲ ﻣﻘﺪار اﻟﺘﺮاﺑﻂ ﺑﻴﻦ ‪ Z t‬و ‪ Z t −k‬ﺑﻌﺪ إزاﻟﺖ ﺗﺄﺛﻴﺮ اﻟﺘﺮاﺑﻂ اﻟﻨﺎﺗﺞ ﻣﻦ اﻟﻤﺘﻐﻴﺮات‬ ‫‪ Z t −1 , Z t −2 ,..., Z t −k +1‬اﻟﻮاﻗﻌﺔ ﺑﻴﻨﻬﻤﺎ وﻳﺮﻣﺰ ﻟﻬﺎ ﻋﻨﺪ اﻟﺘﺨﻠﻒ ‪ k‬ﺑﺎﻟﺮﻣﺰ ‪ φkk‬وأﺣﺪ ﻃﺮق‬ ‫ﺣﺴﺎﺑﻬﺎ ﺗﻘﻮم ﻋﻠﻲ ﺣﺴﺎب ﻣﻌﺎﻣﻞ اﻹﻧﺤﺪار اﻟﺠﺰﺋﻲ ‪ φkk‬ﻓﻲ اﻟﺘﻤﺜﻴﻞ‪:‬‬ ‫‪Z t = φk 1Z t −1 + φk 2 Z t −2 + L + φkk Z t −k + at‬‬

‫ﺣﺴﺎب ‪: φ11‬‬

‫‪Z t = φ11Z t −1 + at‬‬

‫ﺑﻀﺮب ﻃﺮﻓﻲ اﻟﻌﻼﻗﺔ ﺑـ ‪ Z t −1‬وأﺧﺬ اﻟﺘﻮﻗﻊ ﻧﺠﺪ‬

‫) ‪E ( Z t −1Z t ) = φ11 E ( Z t −1Z t −1 ) + E ( Z t −1at‬‬

‫أي‬

‫‪γ 1 = φ11γ 0‬‬ ‫ﺣﻴﺚ ‪ ) E ( Z t −1at ) = 0‬ﺑﺸﻜﻞ ﻋﺎم ‪ E ( Z t −k at ) = 0, k = 1, 2,...‬آﻤﺎ ﺳﻨﺒﻴﻦ ﻻﺣﻘﺎ (‬ ‫وﺑﺎﻟﻘﺴﻤﺔ ﻋﻠﻲ ‪ γ 0‬ﻧﺠﺪ‬ ‫‪φ11 = ρ1‬‬

‫ﺗﻌﺮﻳﻒ ‪ :١١‬ﺑﺸﻜﻞ ﻋﺎم ﺗﻌﺮف داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫‪k =0‬‬ ‫‪k =1‬‬

‫‪k = 2,3,...‬‬

‫‪1,‬‬

‫‪ρ1 ,‬‬ ‫‪ρ1‬‬ ‫‪ρ2‬‬

‫‪ρ k −2‬‬ ‫‪ρ k −3‬‬

‫‪M‬‬

‫‪M‬‬

‫‪ρk‬‬ ‫‪,‬‬ ‫‪ρ k −1‬‬ ‫‪ρ k −2‬‬ ‫‪M‬‬ ‫‪1‬‬

‫ﺣﻴﺚ‬

‫‪ρ1‬‬ ‫‪ρ k −2‬‬ ‫‪ρ k −3‬‬ ‫‪M‬‬

‫‪ρ1‬‬

‫‪L‬‬ ‫‪L‬‬ ‫‪L‬‬ ‫‪L‬‬ ‫‪L‬‬ ‫‪L‬‬ ‫‪L‬‬ ‫‪L‬‬

‫‪ρ1‬‬

‫‪1‬‬

‫‪1‬‬ ‫‪M‬‬

‫‪ρ1‬‬ ‫‪M‬‬

‫‪ρ k −1 ρ k −2‬‬ ‫‪1‬‬ ‫‪ρ1‬‬ ‫‪1‬‬ ‫‪ρ1‬‬ ‫‪M‬‬

‫‪M‬‬

‫‪ρ k −1 ρ k −2‬‬

‫⎧‬ ‫⎪‬ ‫⎪‬ ‫⎪‬ ‫⎪‬ ‫⎪‬ ‫⎪‬ ‫⎪‬ ‫⎨ = ‪φkk‬‬ ‫⎪‬ ‫⎪‬ ‫⎪‬ ‫⎪‬ ‫⎪‬ ‫⎪‬ ‫⎪‬ ‫⎩⎪‬

‫ﺗﺮﻣﺰ اﻟﻲ ﻣﺤﺪدة ﻣﺼﻔﻮﻓﺔ‬

‫اﻟﺘﻌﺮﻳﻒ اﻟﺴﺎﺑﻖ ﺻﻌﺐ اﻹﺳﺘﺨﺪام ﻟﻘﻴﻢ ‪ k‬اﻟﻜﺒﻴﺮة وﻟﻬﺬا ﺳﻮف ﻧﻌﻄﻲ ﺗﻌﺮﻳﻒ ﺁﺧﺮ ﻟﺤﺴﺎب داﻟﺔ‬ ‫اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﺗﻜﺮارﻳﺎ‪:‬‬ ‫‪١٢‬‬

‫ﺗﻌﺮﻳﻒ ‪ ١١‬ب‪ :‬ﺗﺤﺴﺐ ‪ φkk‬ﺗﻜﺮارﻳﺎ ﻣﻦ اﻟﻌﻼﻗﺎت‬

‫‪φ00 = 1, by definition‬‬ ‫‪φ11 = ρ1‬‬ ‫‪k −1‬‬

‫‪, k = 2,3,...‬‬

‫‪ρ k − ∑φk −1, j ρ k − j‬‬ ‫‪j =1‬‬ ‫‪k −1‬‬

‫‪1 − ∑φk −1, j ρ j‬‬

‫= ‪φkk‬‬

‫‪j =1‬‬

‫ﺣﻴﺚ‬ ‫‪φkj = φk −1, j − φkkφk −1,k −1 ,‬‬

‫‪j = 1, 2,..., k − 1‬‬ ‫ﺣﺴﺎب ‪: φ22‬‬ ‫ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ١١‬ب‪:‬‬

‫‪ρ 2 − φ11 ρ1 ρ 2 − ρ12‬‬ ‫= ‪φ22‬‬ ‫=‬ ‫‪1 − φ11 ρ1‬‬ ‫‪1 − ρ12‬‬

‫وذﻟﻚ ﻷن ‪. φ11 = ρ1‬‬

‫ﻣﺜﺎل ‪ :٣‬ﺳﻮف ﻧﺸﺘﻖ اﻵن داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻌﻤﻠﻴﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‪:‬‬ ‫ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ١١‬ب‬ ‫‪φ00 = 1, by definition‬‬ ‫‪φ11 = ρ1 = 0‬‬ ‫وذﻟﻚ ﻣﻦ ﻣﺜﺎل ‪ ١‬اﻟﺴﺎﺑﻖ‬ ‫وﺑﺎﻟﺘﻌﻮﻳﺾ ﻓﻲ ﺗﻌﺮﻳﻒ ‪ ١١‬ب ﻋﻦ ‪ φkk‬ﻧﺠﺪ‬ ‫‪φ22 = φ33 = L = 0‬‬ ‫وهﻜﺬا‪:‬‬ ‫‪⎧1, k = 0‬‬ ‫⎨ = ‪φkk‬‬ ‫‪⎩0, k ≠ 0‬‬ ‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‪:‬‬ ‫‪Partial Autocorrelation function of White Noise‬‬ ‫‪1.0‬‬

‫‪0.0‬‬

‫‪PACF‬‬

‫‪0.5‬‬

‫ﻣﻼﺣﻈﺔ‪ :‬ﻻﺣﻆ أن آﻞ ﻣﻦ داﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻌﻤﻠﻴﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‬ ‫‪Lag‬‬ ‫ﺗﺴﺎوي اﻟﺼﻔﺮ ﻣﻦ اﻟﺘﺨﻠﻒ اﻷول‪ .‬وهﺬﻩ ﺧﺎﺻﻴﺔ ﺟﻤﻴﻊ اﻟﻤﺘﻐﻴﺮات اﻟﻌﺸﻮاﺋﻴﺔ ﻏﻴﺮ اﻟﻤﺘﺮاﺑﻄﺔ او‬ ‫اﻟﻤﺴﺘﻘﻠﺔ‪ .‬ﻹﺧﺘﺒﺎر ﻋﺪم اﻟﺘﺮاﺑﻂ ﺑﻴﻦ ﻗﻴﻢ ﻣﺸﺎهﺪة ﻟﻤﺘﻐﻴﺮ ﻋﺸﻮاﺋﻲ ﺗﺴﺘﺨﺪم داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﺬﻟﻚ‪.‬‬ ‫‪9‬‬

‫‪8‬‬

‫‪7‬‬

‫‪6‬‬

‫‪5‬‬

‫‪4‬‬

‫‪١٣‬‬

‫‪3‬‬

‫‪2‬‬

‫‪1‬‬

‫‪0‬‬

‫ﺗﻌﺮﻳﻒ ‪ : ١٢‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﻌﻴﻨﺔ ‪Sample Autocorrelation Function SACF‬‬ ‫ﻟﻤﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺸﺎهﺪة ‪ z1 , z2 ,L , zn −1 , zn‬وﻳﺮﻣﺰ ﻟﻬﺎ ﺑﺎﻟﺮﻣﺰ ‪ rk , k = 0,1, 2,...‬وﺗﻌﻄﻰ‬

‫ﺑﺎﻟﻌﻼﻗﺔ‪:‬‬

‫‪, k = 0,1, 2, ...‬‬

‫) ‪− z )( zt +k − z‬‬ ‫‪2‬‬

‫)‪−z‬‬

‫‪n −k‬‬

‫‪t‬‬

‫‪∑( z‬‬ ‫‪t =1‬‬

‫‪n‬‬

‫‪t‬‬

‫‪∑( z‬‬

‫= ‪rk‬‬

‫‪t =1‬‬

‫‪1 n‬‬ ‫ﺣﻴﺚ ‪z = ∑ zt‬‬ ‫‪n t =1‬‬ ‫وهﻲ ُﻣﻘ ﱢﺪر ‪ Estimator‬ﻟﺪاﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ أي ‪ ρˆ k = rk , k = 0,1, 2,...‬وﺑﻤﺎ اﻧﻬﺎ ُﻣﻘ ﱢﺪر‬ ‫ﻓﻬﻲ إذا ﺗﺘﻐﻴﺮ ﻋﺸﻮاﺋﻴﺎ ﻣﻦ ﻋﻴﻨﺔ ﻻﺧﺮى وﻟﻬﺬا ﻓﺈن ﻟﻬﺎ اﻟﺨﻮاص اﻟﻌﻴﻨﻴﺔ اﻟﺘﺎﻟﻴﺔ‪:‬‬ ‫‪ -١‬إذا آﺎﻧﺖ ‪ ρ k = 0, k > q‬ﻓﺈن‬ ‫‪q‬‬ ‫⎛‪1‬‬ ‫⎞‬ ‫‪V ( rk ) ≅ ⎜ 1 + 2∑ ρ k2 ⎟ , k > q‬‬ ‫⎝‪n‬‬ ‫‪k =1‬‬ ‫⎠‬ ‫‪1‬‬ ‫وﻓﻲ اﻟﺤﺎﻟﺔ اﻟﺨﺎﺻﺔ ﻋﻨﺪﻣﺎ ‪ ρ k = 0, k > 0‬ﻓﺈن ‪V ( rk ) ≅ , k > 0‬‬ ‫‪n‬‬ ‫‪ -٢‬ﻟﻘﻴﻢ ‪ n‬اﻟﻜﺒﻴﺮة و ‪ ρ k = 0‬ﻓﺈن ‪ rk‬ﻳﻜﻮن ﻟﻬﺎ ﺗﻘﺮﻳﺒﺎ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ وﺑﺎﻟﺘﺎﻟﻲ ﻧﺴﺘﻄﻴﻊ اﻟﻘﻴﺎم‬ ‫ﺑﺎﻻﺧﺘﺒﺎر اﻟﺘﺎﻟﻲ‪:‬‬

‫‪H 0 : ρk = 0‬‬ ‫‪H1 : ρ k ≠ 0‬‬ ‫وذﻟﻚ ﺑﺈﺳﺘﺨﺪام اﻹﺣﺼﺎﺋﺔ‪:‬‬ ‫‪= n rk‬‬

‫‪rk‬‬ ‫‪− 12‬‬

‫‪n‬‬

‫وذﻟﻚ ﻋﻨﺪ ﻣﺴﺘﻮى ﻣﻌﻨﻮﻳﺔ ‪ α = 0.05‬وﺗﺮﻓﺾ ‪ H 0‬إذا آﺎﻧﺖ ‪n rk > 1.96‬‬ ‫‪ -٣‬ﺗﺤﺖ اﻟﻔﺮﺿﻴﺔ ‪ H 0 : ρ k = 0, ∀k‬ﻓﺈن ‪corr ( rk , rk − s ) ≅ 0, s ≠ 0‬‬ ‫‪ُ -٤‬ﺗﻘ ﱠﺪر اﻟﺘﺒﺎﻳﻨﺎت ﻟﺪاﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﻌﻴﻨﺔ آﺎﻟﺘﺎﻟﻲ‪:‬‬

‫⎛‪1‬‬ ‫⎞‬ ‫‪Vˆ ( rk ) ≅ ⎜ 1 + 2∑ rk2 ⎟ , k > q‬‬ ‫⎝‪n‬‬ ‫‪k =1‬‬ ‫⎠‬ ‫ﺗﻌﺮﻳﻒ ‪ :١٣‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻌﻴﻨﺔ ‪Sample Partial Autocorrelation‬‬ ‫‪ Function SPACF‬ﻟﻤﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺸﺎهﺪة ‪ z1 , z2 ,L , zn −1 , zn‬وﻳﺮﻣﺰ ﻟﻬﺎ ﺑﺎﻟﺮﻣﺰ‬ ‫‪ rkk , k = 0,1, 2,...‬ﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‪:‬‬ ‫‪q‬‬

‫‪١٤‬‬

‫‪k =0‬‬ ‫‪k =1‬‬ ‫‪rk −2 r1‬‬ ‫‪rk −3 r2‬‬ ‫‪M‬‬ ‫‪M‬‬ ‫‪r1 rk‬‬ ‫‪,‬‬ ‫‪rk −2 rk −1‬‬ ‫‪rk −3 rk −2‬‬ ‫‪M‬‬ ‫‪M‬‬ ‫‪r1‬‬ ‫‪1‬‬

‫‪k = 2,3,...‬‬

‫‪L‬‬ ‫‪L‬‬ ‫‪L‬‬ ‫‪L‬‬ ‫‪L‬‬ ‫‪L‬‬ ‫‪L‬‬ ‫‪L‬‬

‫‪⎧ 1,‬‬ ‫‪⎪ r,‬‬ ‫‪⎪ 1‬‬ ‫‪⎪ 1‬‬ ‫‪r1‬‬ ‫⎪‬ ‫‪1‬‬ ‫‪⎪ r1‬‬ ‫‪⎪ M‬‬ ‫‪M‬‬ ‫⎪‬ ‫‪rkk = ⎨ rk −1 rk −2‬‬ ‫‪⎪ 1‬‬ ‫‪r1‬‬ ‫⎪‬ ‫‪1‬‬ ‫‪⎪ r1‬‬ ‫‪⎪ M‬‬ ‫‪M‬‬ ‫⎪‬ ‫‪⎪ rk −1 rk −2‬‬ ‫⎪‬ ‫⎩⎪‬

‫و ﻟﺤﺴﺎب داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻌﻴﻨﺔ ﺗﻜﺮارﻳﺎ‪:‬‬

‫ﺗﻌﺮﻳﻒ ‪ ١٣‬ب‪ :‬ﺗﺤﺴﺐ ‪ rkk‬ﺗﻜﺮارﻳﺎ ﻣﻦ اﻟﻌﻼﻗﺎت‬

‫‪r00 = 1, by definition‬‬ ‫‪r11 = r1‬‬ ‫‪k −1‬‬

‫‪, k = 2,3, ...‬‬

‫‪rk − ∑ rk −1, j rk − j‬‬ ‫‪j =1‬‬ ‫‪k −1‬‬

‫‪1 − ∑ rk −1, j rj‬‬

‫= ‪rkk‬‬

‫‪j =1‬‬

‫‪j = 1, 2,..., k − 1‬‬

‫ﺣﻴﺚ‬

‫‪rkj = rk −1, j − rkk rk −1,k −1 ,‬‬

‫وهﻲ اﻳﻀﺎ ﻣﻘﺪﱠر ‪ Estimator‬ﻟﺪاﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻌﻴﻨﺔ أي‬ ‫‪ φˆkk = rkk , k = 0,1, 2,...‬وﺑﻤﺎ اﻧﻬﺎ ُﻣﻘ ﱢﺪر ﻓﻬﻲ إذا ﺗﺘﻐﻴﺮ ﻋﺸﻮاﺋﻴﺎ ﻣﻦ ﻋﻴﻨﺔ ﻻﺧﺮى وﻟﻬﺬا ﻓﺈن‬ ‫ﻟﻬﺎ اﻟﺨﻮاص اﻟﻌﻴﻨﻴﺔ اﻟﺘﺎﻟﻴﺔ‪:‬‬ ‫‪1‬‬ ‫‪V ( rkk ) ≅ , k > 0 -١‬‬ ‫‪n‬‬ ‫‪ -٢‬ﻟﻘﻴﻢ ‪ n‬اﻟﻜﺒﻴﺮة ﻓﺈن ‪ rkk‬ﻳﻜﻮن ﻟﻬﺎ ﺗﻘﺮﻳﺒﺎ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ وﺑﺎﻟﺘﺎﻟﻲ ﻧﺴﺘﻄﻴﻊ اﻟﻘﻴﺎم ﺑﺎﻻﺧﺘﺒﺎر‬ ‫اﻟﺘﺎﻟﻲ‪:‬‬

‫‪H 0 : φkk = 0‬‬ ‫‪H1 : φkk ≠ 0‬‬ ‫وذﻟﻚ ﺑﺈﺳﺘﺨﺪام اﻹﺣﺼﺎﺋﺔ‪:‬‬ ‫‪١٥‬‬

‫‪= n rkk‬‬

‫‪rkk‬‬ ‫‪− 12‬‬

‫‪n‬‬

‫وذﻟﻚ ﻋﻨﺪ ﻣﺴﺘﻮى ﻣﻌﻨﻮﻳﺔ ‪ α = 0.05‬وﺗﺮﻓﺾ ‪ H 0‬إذا آﺎﻧﺖ ‪n rkk > 1.96‬‬

‫‪ -٣‬ﺗﺤﺖ اﻟﻔﺮﺿﻴﺔ ‪ H 0 : φkk = 0, ∀k‬ﻓﺈن ‪corr (φkk ,φk − s ,k −s ) ≅ 0, s ≠ 0‬‬

‫‪ُ -٤‬ﺗﻘ ّﺪر اﻟﺘﺒﺎﻳﻨﺎت ﻟﺪاﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﻌﻴﻨﺔ آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫‪1‬‬ ‫‪Vˆ ( rkk ) ≅ , k > 0‬‬ ‫‪n‬‬ ‫ﻣﺜﺎل ‪ :٤‬اﻟﺒﻴﺎﻧﺎت اﻟﺘﺎﻟﻴﺔ ﺗﻤﺜﻞ اﻟﻄﻠﺐ ﻋﻠﻲ ﻣﻨﺘﺞ ﻣﻌﻴﻦ ﻳﻮﻣﻴﺎ‪:‬‬ ‫‪158 222 248 216 226 239 206 178 169‬‬ ‫أﺣﺴﺐ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻌﻴﻨﺔ وارﺳﻤﻬﻤﺎ‪:‬‬ ‫‪1‬‬ ‫‪1‬‬ ‫اوﻻ‪ :‬ﻧﺤﺴﺐ اﻟﻤﺘﻮﺳﻂ ‪zt = (158 + 222 + L + 169 ) = 206.89‬‬ ‫∑‬ ‫‪9‬‬ ‫‪n t =1‬‬ ‫ﺛﺎﻧﻴﺎ‪ :‬ﻧﺤﺴﺐ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻣﻦ اﻟﻌﻼﻗﺔ‬ ‫‪n‬‬

‫‪, k = 0,1,2,...‬‬

‫=‪z‬‬ ‫‪n −k‬‬

‫) ‪− z )( zt +k − z‬‬ ‫‪2‬‬

‫)‪−z‬‬

‫‪t‬‬

‫‪∑( z‬‬

‫‪n‬‬

‫‪t‬‬

‫‪∑( z‬‬

‫‪t =1‬‬

‫= ‪rk‬‬

‫‪t =1‬‬

‫) ‪(158 × 222 + 222 × 248 + L + 178 × 169‬‬ ‫‪= 0.265116‬‬ ‫) ‪(158 × 158 + 222 × 222 + L + 169 × 169‬‬ ‫) ‪(158 × 248 + 222 × 216 + L + 206 × 169‬‬ ‫= ‪r2‬‬ ‫‪= -0.212‬‬ ‫) ‪(158 × 158 + 222 × 222 + L + 169 × 169‬‬ ‫) ‪(158 × 216 + 222 × 226 + L + 239 × 169‬‬ ‫= ‪r3‬‬ ‫‪= −0.076‬‬ ‫) ‪(158 × 158 + 222 × 222 + L + 169 × 169‬‬ ‫= ‪r1‬‬

‫وهﻜﺬا ‪r8 = 0.230, r7 = 0.104, r6 = −0.242, r5 = −0.387, r4 = −0.183‬‬ ‫ﺛﺎﻟﺜﺎ‪ :‬ﻧﺤﺴﺐ اﻟﺘﺒﺎﻳﻨﺎت ﻣﻦ‬ ‫‪q‬‬ ‫⎛‪1‬‬ ‫⎞‬ ‫‪Vˆ ( rk ) ≅ ⎜ 1 + 2∑ rk2 ⎟ , k > q‬‬ ‫⎝‪n‬‬ ‫‪k =1‬‬ ‫⎠‬ ‫‪1‬‬ ‫≅ ) ‪Vˆ ( r1‬‬ ‫‪9‬‬ ‫‪1‬‬ ‫‪1‬‬ ‫‪2‬‬ ‫‪Vˆ ( r2 ) ≅ (1 + 2 r12 ) = 1 + 2 ( 0.265) = 0.1267‬‬ ‫‪n‬‬ ‫‪9‬‬ ‫‪1‬‬ ‫‪1‬‬ ‫‪2‬‬ ‫‪2‬‬ ‫‪Vˆ ( r3 ) ≅ 1 + 2 ( r12 + r22 ) = 1 + 2 ( 0.265) + ( −0.212 ) = 0.1367‬‬ ‫‪n‬‬ ‫‪9‬‬

‫)‬

‫))‬

‫(‬

‫(‬

‫‪١٦‬‬

‫( )‬

‫(‬

‫‪Vˆ ( r4 ) ≅ 0.138 Vˆ ( r5 ) ≅ 0.1454 Vˆ ( r6 ) ≅ 0.1787‬‬ ‫اﻟﺦ…‬ ‫راﺑﻌﺎ‪ :‬ﻧﺤﺴﺐ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻌﻴﻨﺔ‪:‬‬

‫‪r00 = 1, by definition‬‬ ‫‪r11 = r1 = 0.265‬‬

‫ﺛﻢ ﻧﺤﺴﺐ ﺑﺎﻗﻲ اﻟﺘﺮاﺑﻄﺎت ﻣﻦ اﻟﻌﻼﻗﺎت اﻟﺘﻜﺮارﻳﺔ‬ ‫‪k −1‬‬

‫‪, k = 2,3, ...‬‬

‫‪rk − ∑ rk −1, j rk − j‬‬ ‫‪j =1‬‬ ‫‪k −1‬‬

‫‪1 − ∑ rk −1, j rj‬‬

‫= ‪rkk‬‬

‫‪j =1‬‬

‫ﺣﻴﺚ‬

‫‪j = 1, 2,..., k − 1‬‬

‫‪rkj = rk −1, j − rkk rk −1,k −1 ,‬‬ ‫‪1‬‬

‫‪r2 − r11r1 ( −0.212 ) − ( 0.265)( 0.265) −0.282225‬‬ ‫=‬ ‫=‬ ‫‪1 − r11r1‬‬ ‫)‪1 − ( 0.265)( 0.265‬‬ ‫‪0.929775‬‬

‫=‬

‫‪r2 − ∑ r1, j r2− j‬‬ ‫‪j =1‬‬ ‫‪1‬‬

‫‪1 − ∑ r1, j rj‬‬

‫= ‪r22‬‬

‫‪j =1‬‬

‫‪= −0.30354‬‬ ‫ﻟﺤﺴﺎب ‪ r33‬ﻧﺤﺘﺎج اﻟﻰ ‪ r21‬وﺗﺤﺴﺐ ﻣﻦ‬ ‫‪r21 = r11 − r22 r11 = 0.265 − ( −0.303)( 0.265) = 0.345295‬‬

‫‪2‬‬

‫) ‪r3 − ( r21r2 + r22 r1‬‬ ‫) ‪1 − ( r21r1 + r22 r2‬‬

‫=‬

‫‪r3 − ∑ r2, j r3− j‬‬ ‫‪j =1‬‬ ‫‪k −1‬‬

‫‪1 − ∑ r2, j rj‬‬

‫= ‪r33‬‬

‫‪j=1‬‬

‫) )‪( −0.076 ) − ( ( 0.345)( −0.212 ) + ( −0.303)( 0.265‬‬ ‫) ) ‪1 − ( ( 0.345)( 0.265) + ( −0.303)( −0.212‬‬

‫=‬

‫‪= 0.092‬‬

‫وهﻜﺬا ﻧﺤﺴﺐ ﺑﺎﻗﻲ اﻟﺘﺮاﺑﻄﺎت اﻟﺠﺰﺋﻴﺔ ﻟﻠﻌﻴﻨﺔ‬ ‫‪r88 = 0.042, r77 = 0.013, r66 = −0.207, r55 = −0.294, r44 = −0.298‬‬ ‫وﻟﻬﺎ ﺟﻤﻴﻌﺎ اﻟﺘﺒﺎﻳﻨﺎت ﺗﺴﺎوي ﺗﻘﺮﻳﺒﺎ ‪1 = 0.1111‬‬ ‫‪9‬‬ ‫ﺧﺎﻣﺴﺎ‪ :‬رﺳﻢ دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻌﻴﻨﺔ‬ ‫‪Autocorrelation Function for Demand‬‬

‫‪6‬‬

‫‪١٧‬‬

‫‪3‬‬

‫‪4‬‬

‫‪2‬‬

‫‪Corr‬‬

‫‪1‬‬

‫‪Lag‬‬

‫‪LBQ‬‬

‫‪T‬‬

‫‪Corr‬‬

‫‪Lag‬‬

‫‪LBQ‬‬

‫‪T‬‬

‫‪13.66‬‬

‫‪0.52‬‬

‫‪0.23‬‬

‫‪8‬‬

‫‪0.87‬‬ ‫‪1.50‬‬

‫‪1 0.27 0.80‬‬ ‫‪2 -0.21 -0.59‬‬

‫‪Autocorrelation‬‬

‫‪8‬‬

‫‪7‬‬

‫‪5‬‬

‫‪1.0‬‬ ‫‪0.8‬‬ ‫‪0.6‬‬ ‫‪0.4‬‬ ‫‪0.2‬‬ ‫‪0.0‬‬ ‫‪-0.2‬‬ ‫‪-0.4‬‬ ‫‪-0.6‬‬ ‫‪-0.8‬‬ ‫‪-1.0‬‬

Partial Autocorrelation

Partial Autocorrelation Function for Demand 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

1

2

3

4

5

6

Lag PAC

T

Lag PAC

T

0.27 -0.30 0.09 -0.30 -0.29 -0.21 0.01

0.80 -0.91 0.27 -0.89 -0.88 -0.62 0.04

8 0.04

0.13

1 2 3 4 5 6 7

١٨

7

8

‫اﻟﻔﺼﻞ اﻟﺜﺎﻧﻲ‬ ‫ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ_اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ‪Autoregressive-Moving‬‬ ‫‪ Average Models‬وإﺳﺘﺨﺪاﻣﺎﺗﻬﺎ ﻓﻲ اﻟﺘﻨﺒﺆ‪:‬‬ ‫هﻨﺎك ﻋﺎﺋﻠﺔ آﺒﻴﺮة ﻣﻦ اﻟﻨﻤﺎذج اﻟﺘﻲ ﻳﻄﻠﻖ ﻋﻠﻴﻬﺎ ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ_اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك‬ ‫‪ Autoregressive-Moving Average Models‬واﻟﺘﻲ اﺛﺒﺘﺖ اﻷﺑﺤﺎث اﻟﻜﺜﻴﺮة ﻓﻲ ﻣﺨﺘﻠﻒ‬ ‫اﻟﻤﻴﺎدﻳﻦ اﻟﺘﻄﺒﻴﻘﻴﺔ ﻋﻠﻲ ﺗﻔﻮﻗﻬﺎ اﻟﻬﺎﺋﻞ ﻋﻠﻲ اﻟﻄﺮق اﻟﺘﻘﻠﻴﺪﻳﺔ ﻓﻲ اﻟﺘﻨﺒﺆ‪.‬‬

‫ﺗﻌﺮﻳﻒ ‪ :١٤‬ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ_اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ ) ‪ ( p, q‬وﻳﺮﻣﺰ ﻟﻪ‬ ‫) ‪ ARMA ( p, q‬ﻟﻤﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺸﺎهﺪة } ‪ {z1 , z2 ,K, zn−1 , zn‬ﻳﻜﺘﺐ ﻋﻠﻲ اﻟﺸﻜﻞ‪:‬‬ ‫‪zt = δ + φ1 zt −1 + φ2 zt −2 + L + φ p zt − p + at − θ1at −1 − θ 2 at −2 − L − θ q at −q‬‬

‫ﺣﻴﺚ ) ‪ at WN ( 0,σ 2‬ﻣﺘﺴﻠﺴﻠﺔ ﺿﺠﺔ ﺑﻴﻀﺎء و ∞ < ‪ −∞ < δ‬ﻣﻌﻠﻢ ﺛﺎﺑﺖ ﻳﻤﺜﻞ اﻟﻤﺴﺘﻮي‬ ‫و ‪ φ1 ,φ2 ,K,φ p‬هﻲ ﻣﻌﺎﻟﻢ اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ‪ Autoregressive Parameters‬و‬ ‫‪ θ1 ,θ 2 ,K,θ q‬هﻲ ﻣﻌﺎﻟﻢ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ‪Moving Average Operators‬‬

‫ﺳﻮف ﻧﺴﺘﻌﻴﻦ ﺑﺠﺒﺮ اﻟﻌﻤﺎل ‪ Operators Algebra‬ﻟﺘﺒﺴﻴﻂ هﺬﻩ اﻟﻨﻤﺎذج ﻟﻜﻲ ﻳﺴﻬﻞ اﻟﺘﻌﺎﻣﻞ ﻣﻌﻬﺎ‬

‫ﺗﻌﺮﻳﻒ ‪ :١٥‬ﻋﺎﻣﻞ اﻹزاﺣﺔ اﻟﺨﻠﻔﻲ ‪ Backshift Operator‬وﻳﺮﻣﺰ ﻟﻪ ‪ B‬وﻟﻪ اﻟﺨﻮاص‬ ‫اﻟﺘﺎﻟﻴﺔ‪:‬‬ ‫‪1 − Bzt = zt −1‬‬ ‫‪2 − B m zt = B m−1 ( Bzt ) = B m−2 ( B ( Bzt ) ) = L = zt −m‬‬

‫‪3 − Bc = c, c is a constant‬‬ ‫ﺑﺎﻹﺿﺎﻓﺔ اﻟﻲ ﻋﺎﻣﻞ اﻹزاﺣﺔ اﻟﺨﻠﻔﻲ ﺗﻮﺟﺪ ﻋﻤﺎل اﺧﺮي ﻧﺤﺘﺎج اﻟﻴﻬﺎ ﻻﺣﻘﺎ هﻲ‪:‬‬

‫ﺗﻌﺮﻳﻒ ‪ ١٥‬ب‪:‬‬ ‫‪ -١‬ﻋﺎﻣﻞ اﻹزاﺣﺔ اﻷﻣﺎﻣﻲ ‪ Forewardshift Operator‬وﻳﺮﻣﺰ ﻟﻪ ‪ F‬وﻳﻌﺮف آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫‪F = B −1‬‬ ‫‪ -٢‬ﻋﺎﻣﻞ اﻟﺘﻔﺮﻳﻖ ‪ Difference Operator‬وﻳﺮﻣﺰ ﻟﻪ ∇ وﻳﻌﺮف آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫) ‪∇ = (1 − B‬‬ ‫‪ -٣‬ﻋﺎﻣﻞ اﻟﺘﺠﻤﻴﻊ ‪ Sum Operator‬وﻳﺮﻣﺰ ﻟﻪ ‪ S‬وﻳﻌﺮف آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫‪−1‬‬ ‫) ‪S = ∇ −1 = (1 − B‬‬ ‫اﻵن ﻧﻌﻮد اﻟﻲ ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ_اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ ) ‪ ( p, q‬وﻧﻜﺘﺒﻪ ﻋﻠﻲ‬ ‫اﻟﺸﻜﻞ‪:‬‬

‫‪١٩‬‬

zt − φ1 zt −1 − φ2 zt −2 − L − φ p zt − p = δ + at − θ1at −1 − θ 2 at −2 − L − θ q at −q zt − φ1 Bzt − φ2 B 2 zt − L − φ p B p zt = δ + at − θ1 Bat − θ 2 B 2 at − L − θ q B q at

(1 − φ B − φ B 1

2

2

− L − φ p B p ) zt = δ + (1 − θ1 B − θ 2 B 2 − L − θ q B q ) at

‫أو‬

φ p ( B ) zt = δ + θ q ( B ) at

Autoregressive ‫ هﻮ ﻋﺎﻣﻞ اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‬φ p ( B ) = 1 − φ1B − φ2 B 2 − L − φ p B p ‫ﺣﻴﺚ‬ ‫ هﻮ ﻋﺎﻣﻞ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك‬θ q ( B ) = 1 − θ1 B − θ 2 B 2 − L − θ q B q ‫ و‬Operator Moving Average Operator

:‫أﻣﺜﻠﺔ‬ ‫ وﻳﻜﺘﺐ‬ARMA ( 0,0 ) ‫ وﻳﺮﻣﺰ ﻟﻪ‬Constant Mean Model ‫ ﻧﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﺜﺎﺑﺖ‬-١ :‫ﻋﻠﻲ اﻟﺸﻜﻞ‬ φ0 ( B ) zt = δ + θ 0 ( B ) at ‫او‬ (1) zt = δ + (1) at WN ( 0,σ 2 )

z t = δ + a t , at

:‫ وهﻮ ﻋﻠﻲ اﻟﺸﻜﻞ‬ARMA (1,0 ) ≡ AR (1) ‫ ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ﻣﻦ اﻟﺪرﺟﺔ اﻻوﻟﻲ‬-٢ φ1 ( B ) zt = δ + θ 0 ( B ) at

(1 − φ1B ) zt = δ + at

zt = δ + φ1 zt −1 + at , at

WN ( 0,σ 2 )

:‫ وهﻮ ﻋﻠﻲ اﻟﺸﻜﻞ‬ARMA ( 0,1) ≡ MA (1) ‫ ﻧﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ اﻻوﻟﻲ‬-٣ φ0 ( B ) zt = δ + θ1 ( B ) at zt = δ + (1 − θ1 B ) at

zt = δ + at − θ1at −1 , at

WN ( 0,σ 2 )

:‫ وهﻮ ﻋﻠﻲ اﻟﺸﻜﻞ‬ARMA ( 2,0 ) ≡ AR ( 2 ) ‫ ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ﻣﻦ اﻟﺪرﺟﺔ اﻟﺜﺎﻧﻴﺔ‬-٤ φ 2 ( B ) z t = δ + θ 0 ( B ) at

(1 − φ B − φ B ) z 2

1

2

t

= δ + at

zt = δ + φ1 zt −1 + φ2 zt −2 + at , at

WN ( 0,σ 2 )

:‫ وﻧﻜﺘﺒﻪ ﻋﻠﻲ اﻟﺸﻜﻞ‬ARMA (1,1) (١‫و‬١) ‫ ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ_اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ‬-٥

φ1 ( B ) zt = δ + θ1 ( B ) at (1 − φ1B ) zt = δ + (1 − θ1B ) at

zt = δ + φ1 zt −1 + at − θ1at −1 , at

WN ( 0,σ 2 )

٢٠

‫ﺧﺼﺎﺋﺺ ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ_اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك‪:‬‬ ‫ﺳﻮف ﻧﺪرس اﻟﺨﺼﺎﺋﺺ اﻹﺣﺼﺎﺋﻴﺔ اﻟﺘﻲ ﺗﻤﻴﺰ ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ_اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك وﻣﻌﺮﻓﺔ آﻴﻔﻴﺔ اﻟﺘﻌﺮف‬ ‫ﻋﻠﻲ اﺣﺪ هﺬﻩ اﻟﻨﻤﺎذج ﻣﻦ ﻋﻴﻨﺔ ﻣﺸﺎهﺪة وذﻟﻚ ﻟﺘﻌﻴﻴﻦ او ﺗﺤﺪﻳﺪ ﻧﻤﻮذج ﻣﻨﺎﺳﺐ ﻳﺼﻒ اﻟﻤﺸﺎهﺪات‪.‬‬

‫أوﻻ‪ :‬ﻧﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﺜﺎﺑﺖ)‪:ARMA(٠،٠‬‬ ‫وﻳﻜﺘﺐ ﻋﻠﻲ اﻟﺸﻜﻞ‬

‫‪φ0 ( B ) zt = δ + θ 0 ( B ) at‬‬

‫او‬

‫) ‪WN ( 0, σ 2‬‬

‫‪z t = δ + a t , at‬‬

‫ﺳﻮف ﻧﺸﺘﻖ اﻟﺨﻮاص اﻹﺣﺼﺎﺋﻴﺔ ﻟﻬﺬا اﻟﻨﻤﻮذج وذﻟﻚ ﺑﺈﻳﺠﺎد اﻟﺘﻮﻗﻊ )اﻟﻤﺘﻮﺳﻂ( وداﻟﺘﻲ اﻟﺘﺮاﺑﻂ‬ ‫اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫) ‪E ( z t ) = δ + E ( at‬‬ ‫وذﻟﻚ ﻷن ) ‪WN ( 0, σ 2‬‬

‫‪=δ‬‬ ‫‪at‬‬

‫ﺳﻮف ﻧﺮﻣﺰ ﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺴﻠﺴﻠﺔ ) ‪ E ( zt‬ﺑﺎﻟﺮﻣﺰ ‪ µ‬أي ) ‪ µ = E ( zt‬وﺑﺎﻟﺘﺎﻟﻲ ﻳﻜﻮن ‪δ = µ‬‬

‫وﻳﻜﺘﺐ اﻟﻨﻤﻮذج‪:‬‬

‫‪zt − µ = at‬‬ ‫ﻹﺷﺘﻘﺎق داﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻧﻀﺮب ﻃﺮﻓﻲ اﻟﻤﻌﺎدﻟﺔ اﻟﺴﺎﺑﻘﺔ ﻓﻲ‬ ‫‪ zt −k − µ‬وﻧﺄﺧﺬ اﻟﺘﻮﻗﻊ أي‬

‫⎦⎤ ‪E ⎡⎣( zt −k − µ )( zt − µ ) ⎤⎦ = E ⎡⎣( zt −k − µ ) at‬‬

‫وﻟﻜﻦ ‪ E ⎡⎣( zt −k − µ )( zt − µ ) ⎤⎦ = γ k‬ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ٨‬إذا‬ ‫‪γ k = E ⎡⎣( zt −k − µ ) at ⎤⎦ , k = 0, ±1, ±2,L‬‬ ‫وﻧﺤﻞ هﺬﻩ اﻟﻌﻼﻗﺔ ﺗﻜﺮارﻳﺎ‪:‬‬ ‫⎦⎤ ‪k = 0 : γ 0 = E ⎡⎣( zt − µ ) at‬‬ ‫ﻹﻳﺠﺎد اﻟﻄﺮف اﻷﻳﻤﻦ ﻧﻀﺮب ﻃﺮﻓﻲ ‪ zt − µ = at‬ﻓﻲ ‪ at‬وﻧﺄﺧﺬ اﻟﺘﻮﻗﻊ أي‬ ‫وذﻟﻚ ﻷن ) ‪WN ( 0, σ 2‬‬

‫‪E ⎡⎣( zt − µ ) at ⎤⎦ = E ( at at ) = σ 2‬‬

‫‪ at‬إذا‬

‫‪k = 0 : γ 0 = E ⎡⎣( zt − µ ) at ⎤⎦ = σ 2‬‬ ‫‪k = 1: γ 1 = E ⎡⎣( zt −1 − µ ) at ⎤⎦ = 0‬‬

‫وذﻟﻚ ﻷن‬

‫‪zt −1 − µ = at −1‬‬ ‫‪E ⎡⎣( zt −1 − µ ) at ⎤⎦ = E ( at −1at ) = 0‬‬

‫ﻓﻲ اﻟﺤﻘﻴﻘﺔ ﻓﺈن‬

‫‪٢١‬‬

zt −k − µ = at −k , k = 1, 2,K

E ⎡⎣( zt −k − µ ) at ⎦⎤ = E ( at −k at ) = 0, k = 1, 2,K ‫أي‬

:١ ‫ﻗﺎﻋﺪة‬ ⎧σ , k = 0 E ⎡⎣ ( zt −k − µ ) at ⎤⎦ = E ( at −k at ) = ⎨ ⎩ 0, k = 1, 2,.. 2

‫أي‬

γ0 =σ γ k = 0, k = ±1, ±2,K 2

:‫وﺗﻮﺿﻊ ﻋﻠﻲ ﺷﻜﻞ داﻟﻲ‬

⎧σ , k = 0 γk = ⎨ 2

⎩ 0, k ≠ 0

‫ ﻧﺠﺪ‬γ 0 = σ 2 ‫وﺑﺎﻟﻘﺴﻤﺔ ﻋﻠﻲ‬

γ k ⎧1, k = 0 =⎨ γ 0 ⎩0, k ≠ 0 :‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‬ Autocorrelation function of Constant Mean Model 1.0

Autocorr

ρk =

0.5

0.0 0

1

2

3

4

5

6

7

8

9

Lag

‫ ﻧﺠﺪ‬١١ ‫ﻧﺸﺘﻖ اﻵن داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻣﻦ اﻟﺘﻌﺮﻳﻒ‬

٢٢

φ00 = 1, by definition φ11 = ρ1 , by definition φ11 = 0

1

ρ1 1 0 ρ2 0 0 = =0 ρ1 1 0

ρ1

1

1

ρ1

ρ1 ρ2

1

1

φ22 =

φ33 =

ρ1

0 1

1

ρ1 1 0 ρ2 0 1 ρ3 0 0 = ρ2 1 0 ρ1 0 1

ρ1

1

ρ1 ρ1

1

ρ1 ρ2

0 0 0 =0 0 0 0 0 1

M

1

ρ1

ρ1

1 M

M

ρ k −1 ρ k −2 ρ1 1 ρ1 1

φkk =

M

M

ρ k −1 ρ k −2

L L M L L L M L

ρ1 ρ2

1 0 M 0 = 1 0 M 0

M

ρk ρ k −1 ρ k −2 M 1

0 1 M 0 0 1 M 0

L L M L L L M L

0 0 M 0 0 = = 0, k = 2,3,L 0 1 0 M 1 :‫وﺗﻮﺿﻊ ﻋﻠﻲ ﺷﻜﻞ داﻟﻲ‬

⎧1, k = 0 ⎩0, k ≠ 0

φkk = ⎨

Partial Autocorrelation function of Constant Mean Model

:‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‬

PACF

1.0

0.5

0.0 0

1

2

3

4

5

6

7

8

9

Lag ‫اﻟﺒﻴﻀﺎء اﻻ ﻓﻲ ان ﻟﻪ ﻣﺘﻮﺳﻂ ﻏﻴﺮ ﺻﻔﺮي‬ ‫ ﻧﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﺜﺎﺑﺖ ﻻﻳﻔﺘﺮق ﻋﻦ ﻧﻤﻮذج اﻟﻀﺠﺔ‬:‫ﻣﻼﺣﻈﺔ‬

٢٣

‫ﺛﺎﻧﻴﺎ‪ :‬ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ﻣﻦ اﻟﺪرﺟﺔ اﻻوﻟﻲ )‪ARMA(١،٠) = AR(١‬‬ ‫وهﻮ ﻋﻠﻲ اﻟﺸﻜﻞ‪:‬‬ ‫‪φ1 ( B ) zt = δ + θ 0 ( B ) at‬‬

‫‪(1 − φ1B ) zt = δ + at‬‬

‫) ‪WN ( 0,σ 2‬‬

‫‪zt = δ + φ1 zt −1 + at , at‬‬

‫آﺎﻟﻨﻤﻮذج اﻟﺴﺎﺑﻖ ﺳﻮف ﻧﻮﺟﺪ اﻟﺘﻮﻗﻊ )اﻟﻤﺘﻮﺳﻂ( وداﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‪:‬‬

‫‪(1 − φ1B ) zt = δ + at‬‬

‫‪+ (1 − φ1 B ) at‬‬

‫‪δ‬‬

‫‪−1‬‬

‫) ‪(1 − φ1‬‬

‫‪δ‬‬

‫‪−1‬‬ ‫⎤ ‪+ E ⎡(1 − φ1 B ) at‬‬ ‫⎦‬ ‫⎣ ) ‪(1 − φ1‬‬

‫= ‪zt‬‬

‫= ) ‪E ( zt‬‬

‫اﻟﺤﺪ اﻟﺜﺎﻧﻲ ﻓﻲ اﻟﻄﺮف اﻷﻳﻤﻦ هﻮ‬

‫⎤ ⎞ ‪⎡⎛ ∞ j j‬‬ ‫‪−1‬‬ ‫⎡‬ ‫⎤‬ ‫⎥ ‪E (1 − φ1 B ) at = E ⎢⎜ ∑φ1 B ⎟ at‬‬ ‫⎣‬ ‫⎦‬ ‫‪⎢⎣⎝ j =0‬‬ ‫⎦⎥ ⎠‬ ‫ﻹدﺧﺎل اﻟﺘﻮﻗﻊ ﻋﻠﻲ اﻟﻤﺠﻤﻮع اﻟﻶﻧﻬﺎﺋﻲ ﻳﺠﺐ ان ﺗﻜﻮن اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻶﻧﻬﺎﺋﻴﺔ ∞ < ‪B j‬‬

‫‪j‬‬

‫∞‬

‫‪∑φ‬‬

‫‪1‬‬

‫ﻣﺘﻘﺎرﺑﺔ وذﻟﻚ‬

‫‪j =0‬‬

‫ﻳﺘﺤﻘﻖ إذا آﺎﻧﺖ ‪ φ1 < 1‬وذﻟﻚ إذا اﻋﺘﺒﺮﻧﺎ اﻟﻌﺎﻣﻞ ‪ B‬اﻵن ﻳﻠﻌﺐ دور ﻣﺘﻐﻴﺮ ﻣﺮآﺐ ‪Complex‬‬ ‫‪ Variable‬ﻟﻪ اﻟﺸﻜﻞ ‪ B = a + ib‬وﻟﻪ اﻟﻘﻴﺎس ‪ B = 1‬ﻓﻲ اﻟﺤﻘﻴﻘﺔ ﻻﺑﺪ ان ﻧﺘﻄﻠﺐ ان ﺗﻜﻮن ﺟﺰور او‬ ‫اﺻﻔﺎر ‪ (1 − φ1 B ) = 0‬ﺧﺎرج داﺋﺮة اﻟﻮﺣﺪة أي ‪ B > 1‬أي‬ ‫‪1 − φ1 B = 0‬‬ ‫‪1‬‬

‫‪φ1‬‬ ‫‪> 1 ⇒ φ1 < 1‬‬

‫‪1‬‬

‫‪φ1‬‬

‫=‪B‬‬

‫⇒‪B >1‬‬

‫وهﺬا هﻮ ﺷﺮط اﻹﺳﺘﻘﺮار‪ .‬ﻧﻌﻮد اﻟﻲ اﻟﻌﻼﻗﺔ‬

‫⎤ ⎞ ‪⎡⎛ ∞ j‬‬ ‫‪−1‬‬ ‫⎥ ‪E ⎡(1 − φ1 B ) at ⎤ = E ⎢⎜ ∑φ1 B j ⎟ at‬‬ ‫⎣‬ ‫⎦‬ ‫⎥⎦ ⎠‬ ‫‪⎣⎢⎝ j =0‬‬ ‫⎞ ‪⎡⎛ ∞ j j‬‬ ‫⎤‬ ‫⎥ ) ‪= ⎢⎜ ∑φ1 B ⎟ E ( at‬‬ ‫‪⎢⎣⎝ j =0‬‬ ‫⎦⎥‬ ‫⎠‬ ‫‪=0, ∀t‬‬ ‫وﻳﻜﻮن‬

‫‪٢٤‬‬

‫‪δ‬‬

‫) ‪(1 − φ1‬‬ ‫او‬

‫= ) ‪E ( zt‬‬

‫‪δ‬‬

‫=‪µ‬‬

‫) ‪(1 − φ1‬‬ ‫) ‪∴δ = µ (1 − φ1‬‬

‫وﺑﺎﻟﺘﻌﻮﻳﺾ ﻋﻦ ‪ δ‬ﻓﻲ ﺻﻴﻐﺔ اﻟﻨﻤﻮذج ﻧﺠﺪ‬

‫‪zt = δ + φ1 zt −1 + at‬‬

‫‪= µ (1 − φ1 ) + φ1 zt −1 + at‬‬ ‫‪= µ + φ1 ( zt −1 − µ ) + at‬‬ ‫ﻧﻀﺮب ﻃﺮﻓﻲ اﻟﻤﻌﺎدﻟﺔ اﻟﺴﺎﺑﻘﺔ ﻓﻲ ‪ zt −k − µ‬وﻧﺄﺧﺬ اﻟﺘﻮﻗﻊ أي‬

‫‪( zt − µ ) − φ1 ( zt −1 − µ ) = at‬‬

‫‪E ⎡⎣( zt −k − µ )( zt − µ ) ⎤⎦ − φ1E ⎡⎣( zt −k − µ )( zt −1 − µ ) ⎤⎦ = E ⎡⎣( zt −k − µ ) at ⎤⎦ , k = 0, ±1, ±2,L‬‬ ‫أي‬

‫‪γ k − φ1γ k −1 = E ⎡⎣( zt −k − µ ) at ⎤⎦ , k = 0, ±1, ±2,L‬‬

‫وذﻟﻚ ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ٨‬و ﺗﺤﻞ هﺬﻩ اﻟﻌﻼﻗﺔ ﺗﻜﺮارﻳﺎ آﻤﺎ ﻳﻠﻲ‪:‬‬ ‫⎦⎤ ‪k = 0 : γ 0 − φ1γ 1 = E ⎡⎣( zt − µ ) at‬‬ ‫ﻹﻳﺠﺎد اﻟﻄﺮف اﻷﻳﻤﻦ ﻧﻘﻮم ﺑﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫) ‪E ⎡⎣ at ( zt − µ ) ⎤⎦ − φ1 E ⎡⎣ at ( zt −1 − µ ) ⎤⎦ = E ( at at‬‬ ‫‪E ⎡⎣ at ( zt − µ ) ⎤⎦ − φ1 × ( 0 ) = σ 2‬‬ ‫‪∴ E ⎡⎣ at ( zt − µ ) ⎤⎦ = σ 2‬‬ ‫إذا‬

‫‪γ 0 − φ1γ 1 = σ‬‬ ‫‪k = 1: γ 1 − φ1γ 0 = E ⎡⎣( zt −1 − µ ) at ⎤⎦ = 0‬‬ ‫‪2‬‬

‫ﻓﻲ اﻟﺤﻘﻴﻘﺔ‬

‫ﺑﻘﺴﻤﺔ اﻟﻤﻌﺎدﻟﺔ اﻷﺧﻴﺮة ﻋﻠﻲ ‪γ 0‬‬

‫‪γ k − φ1γ k −1 = 0, k = 1, 2,L‬‬ ‫ﻧﺠﺪ‬

‫‪ρ k − φ1 ρ k −1 = 0, k = 1, 2,L‬‬

‫أو‬

‫وﺑﻤﺎ ان ‪ρ 0 = 1‬‬

‫‪ρ k = φ1 ρ k −1 , k = 1, 2,L‬‬ ‫ﻓﺈن‪:‬‬

‫‪ρ1 = φ1 ρ 0 = φ1‬‬ ‫‪ρ 2 = φ1 ρ1 = φ12‬‬

‫‪٢٥‬‬

‫‪M‬‬

‫‪ρ k = φ1k‬‬ ‫أو ﺑﺸﻜﻞ داﻟﺔ‬

‫‪ρ k = φ , k = 0, ±1, ±2,L‬‬ ‫وذﻟﻚ ﻷن ‪ ρ − k = ρ k , ∀k‬ﺳﻮف ﻧﻨﻈﺮ ﻣﻦ اﻵن وﺻﺎﻋﺪا ﻟﻠﺸﻖ اﻟﻤﻮﺟﺐ ﻣﻦ ‪ ρ k‬أي‬ ‫‪k‬‬ ‫‪ρ k = φ1 , k = 0,1, 2,L‬‬ ‫هﺬﻩ اﻟﺪاﻟﺔ ﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‪:‬‬ ‫‪ -١‬ﻋﻨﺪﻣﺎ ﺗﻜﻮن ‪φ1 > 0‬‬ ‫‪k‬‬ ‫‪1‬‬

‫‪Autocorrelation function of AR(1) Model‬‬ ‫‪0.5‬‬ ‫‪0.4‬‬

‫‪ACF‬‬

‫‪0.3‬‬ ‫‪0.2‬‬ ‫‪0.1‬‬ ‫‪0.0‬‬ ‫‪10‬‬

‫‪9‬‬

‫‪8‬‬

‫‪7‬‬

‫‪6‬‬

‫‪4‬‬

‫‪5‬‬

‫‪3‬‬

‫‪2‬‬

‫‪1‬‬

‫‪0‬‬

‫‪Lag‬‬

‫‪ -٢‬ﻋﻨﺪﻣﺎ ﺗﻜﻮن ‪φ1 < 0‬‬ ‫‪Autocorrelation function of AR(1) Model‬‬ ‫‪0.3‬‬ ‫‪0.2‬‬ ‫‪0.1‬‬ ‫‪0.0‬‬

‫‪-0.2‬‬ ‫‪-0.3‬‬ ‫‪-0.4‬‬ ‫‪-0.5‬‬ ‫‪10‬‬

‫‪9‬‬

‫‪8‬‬

‫‪7‬‬

‫‪6‬‬

‫‪4‬‬

‫‪5‬‬

‫‪Lag‬‬

‫ﻧﺸﺘﻖ اﻵن داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬ ‫ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ١١‬ﻧﺠﺪ‬

‫‪٢٦‬‬

‫‪3‬‬

‫‪2‬‬

‫‪1‬‬

‫‪0‬‬

‫‪ACF‬‬

‫‪-0.1‬‬

φ00 = 1, by definition φ11 = ρ1 = φ1 , by definition 1

φ22 =

ρ1 1

ρ1

1 φ1 ρ1 ρ 2 φ1 φ12 0 = = =0 1 φ1 1 − φ12 ρ1 1 φ1 1

M 1

ρ1

ρ1

1 M

M

ρ ρ k −2 φkk = k −1 1 ρ1 1 ρ1 M

M

ρ k −1 ρ k −2

L L L L L L L L

ρ1 ρ2 M

1

φ1

φ1

1 M

M

ρk φ = ρ k −1 1 ρ k −2 φ1

φ1k −2 φ1 L

k −1 1

M 1

φ

L φ1 L φ12 L M L φ1k

M

φ

k −1 1

φ1k L φ1k −1

1 M

L L

k −2 1

=

0 >0

M 1

‫ وﻧﻜﺘﺐ‬φ1 ‫ﻣﺤﺪدة اﻟﺒﺴﻂ ﺗﺴﺎوي ﺻﻔﺮا ﻷن اﻟﻌﺎﻣﻮد اﻷﺧﻴﺮ ﻳﺴﺎوي اﻟﻌﺎﻣﻮد اﻷول ﻣﻀﺮوﺑﺎ ﻓﻲ‬ :‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻋﻠﻲ اﻟﺸﻜﻞ اﻟﺪاﻟﻲ‬ ⎧ 1, k = 0 ⎪ φkk = ⎨φ1 , k = 1 ⎪ 0, k ≥ 2 ⎩ :‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‬ φ1 > 0 ‫ ﻋﻨﺪﻣﺎ ﺗﻜﻮن‬-١ Partial Autocorrelation function of AR(1) Model 0.5

PACF

0.4 0.3 0.2 0.1 0.0 0

1

2

3

4

5

6

7

8

9

10

Lag

φ1 < 0 ‫ ﻋﻨﺪﻣﺎ ﺗﻜﻮن‬-٢

٢٧

‫‪Partial Autocorrelation function of AR(1) Model‬‬ ‫‪0.0‬‬ ‫‪-0.1‬‬

‫‪-0.3‬‬

‫‪PACF‬‬

‫‪-0.2‬‬

‫‪-0.4‬‬ ‫‪-0.5‬‬

‫‪10‬‬

‫‪9‬‬

‫‪8‬‬

‫‪7‬‬

‫‪6‬‬

‫‪4‬‬

‫‪5‬‬

‫‪3‬‬

‫‪2‬‬

‫‪1‬‬

‫‪0‬‬

‫‪Lag‬‬

‫ﻣﻼﺣﻈﺔ‪ :‬داﺋﻤﺎ ﻻﺗﺮﺳﻢ أي ﻣﻦ ‪ ρ 0 = 1‬او ‪ φ00 = 1‬ﻓﻲ اﻷﺷﻜﺎل اﻟﺒﻴﺎﻧﻴﺔ‪.‬‬

‫ﻣﻨﺎﻗﺸﺔ اﻟﻨﻤﻮذج‪:‬‬ ‫‪-١‬‬ ‫‪-٢‬‬ ‫‪-٣‬‬ ‫‪-٤‬‬

‫ﻋﻨﺪﻣﺎ ﺗﻜﻮن ‪) φ1 < 1‬ﺷﺮط اﻹﺳﺘﻘﺮار( ﻓﺈن ) ‪ E ( zt ) = δ (1 − φ1‬وهﻮﺛﺎﺑﺖ ﻟﺠﻤﻴﻊ ﻗﻴﻢ ‪t‬‬ ‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ داﻟﺔ ﻟﻠﺘﺨﻠﻒ ‪ k‬ﻓﻘﻂ وﻻﺗﻌﺘﻤﺪ ﻋﻠﻲ اﻟﺰﻣﻦ ‪t‬‬ ‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﺗﺘﺨﺎﻣﺪ اﺳﻴﺎ ﻓﻲ إﺗﺠﺎﻩ واﺣﺪ إﺑﺘﺪاءا ﻣﻦ ‪ ρ1‬ﻋﻨﺪﻣﺎ ﺗﻜﻮن ‪ φ1 > 0‬وﺗﺘﺨﺎﻣﺪ‬ ‫اﺳﻴﺎ ﻣﺘﺮددة ﺑﻴﻦ اﻟﻘﻴﻢ اﻟﻤﻮﺟﺒﺔ واﻟﺴﺎﻟﺒﺔ ﻋﻨﺪﻣﺎ ﺗﻜﻮن ‪φ1 < 0‬‬ ‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻬﺎ ﻗﻴﻤﺔ واﺣﺪة ﻏﻴﺮ ﺻﻔﺮﻳﺔ ) ﻣﻊ ﻋﺪم اﻟﻨﻈﺮ اﻟﻲ ‪ ( φ00‬وﻳﻜﻮن‬ ‫إﺗﺠﺎهﻬﺎ ﺣﺴﺐ إﺷﺎرة ‪ φ1‬وﻣﻘﺪارهﺎ ﻳﺴﺎوي ‪φ1‬‬

‫ﺛﺎﻟﺜﺎ‪ :‬ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ﻣﻦ اﻟﺪرﺟﺔ اﻟﺜﺎﻧﻴﺔ )‪: ARMA(٢،٠) = AR(٢‬‬ ‫وﻳﻜﺘﺐ ﻋﻠﻲ اﻟﺸﻜﻞ‪:‬‬ ‫‪φ 2 ( B ) zt = δ + θ 0 ( B ) a t‬‬ ‫‪= δ + at‬‬ ‫) ‪WN ( 0,σ 2‬‬

‫‪(1 − φ B + φ B ) z‬‬ ‫‪2‬‬

‫‪t‬‬

‫‪2‬‬

‫‪1‬‬

‫‪zt = δ + φ1 zt −1 + φ2 zt −2 + at , at‬‬

‫آﺎﻟﺴﺎﺑﻖ ﻧﻮﺟﺪ اﻟﻤﺘﻮﺳﻂ وداﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‪:‬‬ ‫‪(1 − φ1B − φ2 B ) zt = δ + at‬‬ ‫‪2‬‬

‫‪−1‬‬ ‫‪δ‬‬ ‫‪+ (1 − φ1 B − φ2 B 2 ) at‬‬ ‫) ‪(1 − φ1 − φ2‬‬ ‫‪−1‬‬ ‫‪δ‬‬ ‫⎤ ‪+ E ⎡(1 − φ1 B − φ2 B 2 ) at‬‬ ‫⎦⎥‬ ‫⎣⎢ ) ‪(1 − φ1 − φ2‬‬

‫‪٢٨‬‬

‫= ‪zt‬‬

‫= ) ‪E ( zt‬‬

‫∞ ⎛‬ ‫⎞‬ ‫اﻟﺤﺪ اﻟﺜﺎﻧﻲ ﻓﻲ اﻟﻄﺮف اﻷﻳﻤﻦ ﻣﺠﻤﻮع ﻻﻧﻬﺎﺋﻲ ﻋﻠﻰ اﻟﺸﻜﻞ ⎟ ‪ E ⎜ ∑ψ j at − j‬وﻟﻜﻲ ﻧﺪﺧﻞ اﻟﺘﻮﻗﻊ‬ ‫‪⎝ j =0‬‬ ‫⎠‬ ‫∞‬

‫داﺧﻞ اﻟﺘﺠﻤﻴﻊ اﻟﻼﻧﻬﺎﺋﻲ ﻻﺑﺪ ان ﺗﻜﻮن‬

‫‪∑ψ a‬‬

‫‪j t− j‬‬

‫ﻣﺘﻘﺎرﺑﺔ ﻓﻲ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺮﺑﻊ وهﺬا ﻳﺘﺤﻘﻖ إذا‬

‫‪j =0‬‬

‫وﻓﻘﻂ إذا آﺎن ∞ <‬

‫‪2‬‬ ‫‪j‬‬

‫∞‬

‫‪∑ψ‬‬

‫وهﺬا ﻳﺘﺤﻘﻖ إذا ﺣﻘﻘﺖ ﻣﻌﺎﻟﻢ اﻹﻧﺤﺪار اﻟﺬاﺗﻲ اﻟﺸﺮوط اﻟﺘﺎﻟﻴﺔ‪:‬‬

‫‪j =0‬‬

‫‪φ2 − φ1 < 1‬‬ ‫‪φ2 + φ1 < 1‬‬ ‫‪−1 < φ2 < 1‬‬ ‫واﻟﺘﻲ ﺗﺴﻤﻲ ﺑﺸﺮوط اﻹﺳﺘﻘﺮار ) هﺬﻩ اﻟﺸﺮوط ﺗﻨﺘﺞ اﻳﻀﺎ ﻣﻦ آﻮن ﺟﺰور او أﺻﻔﺎر‬ ‫‪ (1 − φ1 B − φ2 B 2 ) = 0‬ﺧﺎرج داﺋﺮة اﻟﻮﺣﺪة ( ‪ .‬إذا ﺗﺤﻘﻘﺖ ﺷﺮوط اﻹﺳﺘﻘﺮار ﻓﺈن‬ ‫‪−1‬‬ ‫‪−1‬‬ ‫‪E ⎡(1 − φ1 B − φ2 B 2 ) at ⎤ = ⎡(1 − φ1 B − φ2 B 2 ) E ( at ) ⎤ = 0, ∀t‬‬ ‫⎢⎣ ⎦⎥‬ ‫⎦⎥‬ ‫⎢⎣‬

‫وﻳﻜﻮن‬

‫‪δ‬‬ ‫) ‪(1 − φ1 − φ2‬‬

‫و ﺑﺎﻟﺘﻌﻮﻳﺾ ﻋﻦ ‪ δ‬ﻓﻲ ﺻﻴﻐﺔ اﻟﻨﻤﻮذج ﻧﺠﺪ‬

‫= ) ‪µ = E ( zt‬‬

‫‪δ = (1 − φ1 − φ2 ) µ‬‬ ‫‪zt = (1 − φ1 − φ2 ) µ + φ1 zt −1 + φ2 zt −2 + at‬‬ ‫‪= µ + φ1 ( zt −1 − µ ) + φ2 ( zt −2 − µ ) + at‬‬

‫‪( zt − µ ) − φ1 ( zt −1 − µ ) − φ2 ( zt −2 − µ ) = at‬‬ ‫ﻧﻀﺮب اﻟﻤﻌﺎدﻟﺔ اﻟﺴﺎﺑﻘﺔ ﻓﻲ ‪ zt −k − µ‬وﻧﺄﺧﺬ اﻟﺘﻮﻗﻊ ﻧﺠﺪ‪:‬‬

‫⎦⎤ ) ‪E ⎡⎣( zt − µ )( zt −k − µ ) − φ1 ( zt −1 − µ )( zt −k − µ ) − φ2 ( zt −2 − µ )( zt −k − µ‬‬ ‫‪= E ⎡⎣ at ( zt −k − µ ) ⎤⎦ , k = 0, ±1, ±2,...‬‬

‫أي‬ ‫⎦⎤ ) ‪E ⎡⎣( zt − µ )( zt −k − µ ) ⎦⎤ − φ1 E ⎡⎣ ( zt −1 − µ )( zt −k − µ ) ⎤⎦ − φ2 E ⎡⎣( zt −2 − µ )( zt −k − µ‬‬ ‫‪= E ⎡⎣ at ( zt −k − µ ) ⎤⎦ , k = 0, ±1, ±2,...‬‬

‫أو‬

‫‪γ k − φ1γ k −1 − φ2γ k −2 = E ⎡⎣ at ( zt −k − µ ) ⎤⎦ , k = 0, ±1, ±2,...‬‬

‫وذﻟﻚ ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ٨‬اﻵن ﻧﺤﻞ هﺬﻩ اﻟﻌﻼﻗﺔ ﺗﻜﺮارﻳﺎ آﻤﺎ ﻳﻠﻲ‪:‬‬

‫‪٢٩‬‬

‫‪k = 0 : γ 0 − φ1γ −1 − φ2γ −2 = E ⎡⎣ at ( zt − µ ) ⎤⎦ = σ 2 ⇒ γ 0 = φ1γ 1 − φ2γ 2 + σ 2‬‬ ‫وذﻟﻚ ﻣﻦ ﻗﺎﻋﺪة ‪١‬‬ ‫‪k = 1: γ 1 − φ1γ 0 − φ2γ 1 = 0 ⇒ γ 1 = φ1γ 0 − φ2γ 1‬‬ ‫‪k = 2 : γ 2 − φ1γ 1 − φ2γ 0 = 0 ⇒ γ 2 = φ1γ 1 − φ2γ 0‬‬ ‫وﺑﺸﻜﻞ ﻋﺎم‬ ‫‪k ≥ 1: γ k = φ1γ k −1 + φ2γ k −2‬‬ ‫ﺑﻘﺴﻤﺔ اﻟﻄﺮﻓﻴﻦ ﻋﻠﻲ ‪ γ 0‬ﻧﺠﺪ‬ ‫‪ρ k = φ1 ρ k −1 + φ2 ρ k −2 , k = 1, 2, ...‬‬ ‫) ﻣﻼﺣﻈﺔ‪ :‬ﺑﻮﺿﻊ اﻟﻤﻌﺎدﻟﺔ اﻟﺴﺎﺑﻘﺔ ﻋﻠﻲ اﻟﺸﻜﻞ ‪ρ k − φ1 ρ k −1 − φ2 ρ k −2 = 0, k = 1, 2,...‬‬ ‫ﻧﺠﺪ اﻧﻬﺎ ﻣﻌﺎدﻟﺔ ﻓﺮوﻗﻴﺔ ﻣﻦ اﻟﺪرﺟﺔ اﻟﺜﺎﻧﻴﺔ واﻟﺘﻲ ﻳﻤﻜﻦ ﺣﻠﻬﺎ ﺑﺸﻜﻞ ﻣﻐﻠﻖ ﺑﺈﺳﺘﺨﺪام ﻃﺮق ﺣﻞ‬ ‫اﻟﻤﻌﺎدﻻت اﻟﻔﺮوﻗﻴﺔ وﻟﻜﻦ هﺬا ﺧﺎرج ﻧﻄﺎق اﻟﻤﻘﺮر اﻟﺤﺎﻟﻲ(‬ ‫ﺳﻮف ﻧﺤﻞ اﻟﻌﻼﻗﺔ اﻟﺴﺎﺑﻘﺔ ﺑﺎﻟﻄﺮﻳﻘﺔ اﻟﺘﻜﺮارﻳﺔ واﻟﺘﻲ ﺗﺤﺘﺎج اﻟﻲ ﻗﻴﻤﺘﻴﻦ اوﻟﻴﺘﻴﻦ‪:‬‬ ‫‪1 − ρ0 = 1‬‬ ‫‪φ‬‬ ‫‪2 − ρ1 = φ1 ρ 0 + φ2 ρ −1 ⇒ ρ1 = 1‬‬ ‫‪1 − φ2‬‬ ‫وﻣﻨﻬﺎ ﻧﺠﺪ‬ ‫‪2‬‬ ‫‪φ‬‬ ‫‪ρ 2 = φ1 ρ1 + φ2 ρ0 ⇒ ρ 2 = 1 + φ2‬‬ ‫‪1 − φ2‬‬ ‫وهﻜﺬا اﻟﺦ…‬ ‫اﻷﺷﻜﺎل اﻟﺘﺎﻟﻴﺔ هﻲ ﻟﺪوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻌﻤﻠﻴﺔ ) ‪AR ( 2‬‬ ‫‪ -١‬اﻟﺸﻜﻞ )‪φ1 = 0.4, φ2 = 0.4 (١‬‬ ‫‪ -٢‬اﻟﺸﻜﻞ )‪φ1 = 1.5, φ2 = −0.8 (٢‬‬ ‫‪ -٣‬اﻟﺸﻜﻞ )‪φ1 = 0.5, φ2 = −0.6 (٣‬‬ ‫ﺷﻜﻞ )‪(١‬‬ ‫‪ACF‬‬ ‫‪0.7‬‬ ‫‪0.6‬‬ ‫‪0.5‬‬

‫‪0.3‬‬ ‫‪0.2‬‬ ‫‪0.1‬‬ ‫‪0.0‬‬ ‫‪20‬‬

‫‪10‬‬

‫‪Lag‬‬

‫‪٣٠‬‬

‫‪0‬‬

‫‪ACF‬‬

‫‪0.4‬‬

(٢) ‫ﺷﻜﻞ‬ ACF 1.0

ACF

0.5

0.0

-0.5

0

10

20

Lag

(٣) ‫ﺷﻜﻞ‬ ACF

ACF

0.5

0.0

-0.5

0

10

20

Lag

:‫ آﺎﻟﺘﺎﻟﻲ‬AR ( 2 ) ‫اﻵن ﻧﺸﺘﻖ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻌﻤﻠﻴﺔ‬

φ00 = 1, by definition φ11 = ρ1 , by definition 1 ρ1 ρ1 ρ 2 ρ 2 − ρ12 = ≠0 φ22 = 1 ρ1 1 − ρ12 ρ1 1

٣١

‫‪=0‬‬

‫‪ρ1 = φ1 + φ2 ρ1‬‬ ‫‪ρ 2 = φ1 ρ1 + φ2‬‬ ‫‪ρ3 = φ1 ρ 2 + φ2 ρ1‬‬

‫‪ρ1‬‬ ‫‪1‬‬

‫‪ρ1‬‬

‫‪>0‬‬

‫‪ρ1‬‬ ‫‪1‬‬ ‫‪ρ2‬‬ ‫‪ρ1‬‬ ‫‪ρ3 ρ 2‬‬ ‫=‬ ‫‪ρ2‬‬ ‫‪ρ1‬‬

‫‪ρ1‬‬

‫‪1‬‬

‫‪1‬‬

‫‪ρ1‬‬ ‫‪ρ2‬‬

‫‪ρ1‬‬ ‫‪ρ1‬‬ ‫‪1‬‬

‫‪ρ1‬‬

‫‪1‬‬

‫= ‪φ33‬‬

‫‪ρ1‬‬ ‫‪ρ2‬‬

‫‪1‬‬ ‫وذﻟﻚ ﻷن اﻟﻌﻤﻮد اﻷﺧﻴﺮ ﻓﻲ ﻣﺤﺪدة اﻟﺒﺴﻂ هﻮ ﺗﺮآﻴﺐ ﺧﻄﻲ ﻣﻦ اﻟﻌﻤﻮدﻳﻦ اﻷول واﻟﺜﺎﻧﻲ‪ ،‬آﺬﻟﻚ‬ ‫‪1‬‬ ‫‪ρ1 L ρ1‬‬ ‫‪1‬‬ ‫‪ρ1 L‬‬ ‫‪ρ1 = φ1 ρ 0 + φ2 ρ1‬‬ ‫‪ρ1‬‬ ‫‪1 L ρ2‬‬ ‫‪ρ1‬‬ ‫‪1 L‬‬ ‫‪ρ 2 = φ1 ρ1 + φ2 ρ 0‬‬ ‫‪M‬‬ ‫‪M‬‬ ‫‪L M‬‬ ‫‪M‬‬ ‫‪M‬‬ ‫‪L‬‬ ‫‪M‬‬ ‫‪ρ‬‬ ‫‪ρ k −2 L ρ k‬‬ ‫‪ρ‬‬ ‫‪ρ k −2 L ρ k = φ1 ρ k −1 + φ2 ρ k −2‬‬ ‫‪φkk = k −1‬‬ ‫‪= k −1‬‬ ‫‪= 0, k = 3, 4,...‬‬ ‫‪1‬‬ ‫‪ρ1 L ρ k −1‬‬ ‫‪>0‬‬ ‫‪1 L ρ k −2‬‬ ‫‪ρ1‬‬ ‫‪M‬‬ ‫‪M‬‬ ‫‪L‬‬ ‫‪M‬‬ ‫‪ρ k −1 ρ k −2 L 1‬‬ ‫وذﻟﻚ اﻳﻀﺎ ﻟﻨﻔﺲ اﻟﺴﺒﺐ اﻟﺴﺎﺑﻖ‪ .‬إذا‬

‫‪k =0‬‬ ‫‪⎧ 1,‬‬ ‫‪⎪ ρ,‬‬ ‫‪k =1‬‬ ‫‪1‬‬ ‫⎪⎪‬ ‫‪φkk = ⎨ ρ 2 − ρ12‬‬ ‫‪⎪ 1− ρ2 , k = 2‬‬ ‫‪1‬‬ ‫⎪‬ ‫‪k ≥3‬‬ ‫‪⎪⎩ 0,‬‬

‫اﻷﺷﻜﺎل اﻟﺘﺎﻟﻴﺔ هﻲ ﻟﺪوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻌﻤﻠﻴﺔ ) ‪AR ( 2‬‬ ‫‪ -٤‬اﻟﺸﻜﻞ )‪φ1 = 0.4, φ2 = 0.4 (٤‬‬ ‫‪ -٥‬اﻟﺸﻜﻞ )‪φ1 = 1.5, φ2 = −0.8 (٥‬‬ ‫‪ -٦‬اﻟﺸﻜﻞ )‪φ1 = 0.5, φ2 = −0.6 (٦‬‬ ‫ﺷﻜﻞ )‪(٤‬‬

‫‪PACF‬‬ ‫‪0.7‬‬ ‫‪0.6‬‬ ‫‪0.5‬‬

‫‪0.3‬‬ ‫‪0.2‬‬ ‫‪0.1‬‬ ‫‪0.0‬‬ ‫‪20‬‬

‫‪10‬‬

‫‪Lag‬‬

‫‪٣٢‬‬

‫‪0‬‬

‫‪PACF‬‬

‫‪0.4‬‬

‫ﺷﻜﻞ )‪(٥‬‬ ‫‪PACF‬‬ ‫‪1‬‬

‫‪PACF‬‬

‫‪0‬‬

‫‪-1‬‬

‫‪20‬‬

‫‪0‬‬

‫‪10‬‬

‫‪Lag‬‬

‫ﺷﻜﻞ )‪(٦‬‬ ‫‪PACF‬‬ ‫‪0.3‬‬ ‫‪0.2‬‬ ‫‪0.1‬‬

‫‪-0.2‬‬

‫‪PACF‬‬

‫‪0.0‬‬ ‫‪-0.1‬‬

‫‪-0.3‬‬ ‫‪-0.4‬‬ ‫‪-0.5‬‬ ‫‪-0.6‬‬ ‫‪20‬‬

‫‪0‬‬

‫‪10‬‬

‫‪Lag‬‬

‫راﺑﻌﺎ‪ :‬ﻧﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ اﻷوﻟﻲ )‪: ARMA(٠،١) = MA(١‬‬ ‫وﺗﻜﺘﺐ ﻋﻠﻲ اﻟﺸﻜﻞ‪:‬‬ ‫‪φ0 ( B ) zt = δ + θ1 ( B ) at‬‬

‫) ‪WN ( 0,σ 2‬‬

‫‪zt = δ + (1 − θ1 B ) at‬‬ ‫‪zt = δ + at − θ1at −1 , at‬‬

‫اﻵن ﻧﻮﺟﺪ اﻟﻤﺘﻮﺳﻂ وداﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‪:‬‬ ‫‪E ( zt ) = E (δ + at − θ1at −1 ) = δ‬‬ ‫‪∴µ = δ‬‬

‫وﻧﻜﺘﺐ اﻟﻨﻤﻮذج‬ ‫ﺑﻀﺮب هﺬﻩ اﻟﻤﻌﺎدﻟﺔ ﻓﻲ ‪ zt −k − µ‬وأﺧﺬ اﻟﺘﻮﻗﻊ ﻧﺠﺪ‬

‫‪zt − µ = at − θ1at −1‬‬

‫‪E ⎡⎣( zt − µ )( zt −k − µ ) ⎤⎦ = E ⎡⎣( zt −k − µ ) at ⎤⎦ − θ1E ⎡⎣( zt −k − µ ) at −1 ⎤⎦ , k = 0, ±1, ±2,...‬‬

‫‪٣٣‬‬

‫او‬

‫‪γ k = E ⎡⎣( zt −k − µ ) at ⎤⎦ − θ1 E ⎡⎣( zt −k − µ ) at −1 ⎤⎦ , k = 0, ±1, ±2,...‬‬

‫وﺑﺤﻠﻬﺎ ﺗﻜﺮارﻳﺎ‬

‫⎦⎤ ‪k = 0 : γ 0 = E ⎡⎣( zt − µ ) at ⎤⎦ − θ1 E ⎡⎣( zt − µ ) at −1‬‬

‫ﻧﻮﺟﺪ آﻞ ﻣﻦ ⎦⎤ ‪ E ⎡⎣( zt − µ ) at‬و ⎦⎤ ‪ E ⎡⎣( zt − µ ) at −1‬آﺎﻵﺗﻲ‪:‬‬ ‫‪E ⎡⎣( zt − µ ) at ⎤⎦ = E ( at at ) − θ1 E ( at −1at ) = σ 2‬‬ ‫‪E ⎡⎣( zt − µ ) at −1 ⎤⎦ = E ( at at −1 ) − θ1E ( at −1at −1 ) = −θ1σ 2‬‬

‫) ‪∴γ 0 = σ 2 − θ1 ( −θ1σ 2 ) = σ 2 (1 + θ12‬‬

‫⎦⎤ ‪k = 1: γ 1 = E ⎡⎣( zt −1 − µ ) at ⎤⎦ − θ1 E ⎡⎣( zt −1 − µ ) at −1‬‬ ‫‪−θ1‬‬ ‫‪γ1‬‬ ‫=‬ ‫‪γ 0 1 + θ12‬‬ ‫وذﻟﻚ ﺑﺈﺳﺘﺨﺪام اﻟﻘﺎﻋﺪة ‪١‬‬

‫= ‪∴ γ 1 = −θ1σ 2 ⇒ ρ1‬‬

‫⎦⎤ ‪k = 2 : γ 2 = E ⎡⎣( zt −2 − µ ) at ⎤⎦ − θ1 E ⎡⎣( zt −2 − µ ) at −1‬‬ ‫‪∴ γ 2 = 0 ⇒ ρ2 = 0‬‬

‫أﻳﻀﺎ ﻣﻦ ﻗﺎﻋﺪة ‪ ١‬وﺑﺸﻜﻞ ﻋﺎم ﻓﺈن‬

‫‪k ≥ 2 : γ k = 0 ⇒ ρk = 0‬‬ ‫وهﻜﺬا ﻓﺈن داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻨﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ اﻷوﻟﻲ )‪ MA (1‬هﻲ ﻋﻠﻲ‬ ‫اﻟﺸﻜﻞ‪:‬‬ ‫‪⎧ 1,‬‬ ‫‪k =0‬‬ ‫⎪‬ ‫‪⎪ −θ‬‬ ‫‪ρk = ⎨ 1 2 , k = 1‬‬ ‫‪⎪ 1 + θ1‬‬ ‫‪⎪⎩ 0‬‬ ‫‪k ≥2‬‬ ‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‪:‬‬ ‫‪ -١‬ﻋﻨﺪﻣﺎ ‪θ1 = 0.8‬‬ ‫‪ACF‬‬ ‫‪0.0‬‬ ‫‪-0.1‬‬

‫‪-0.3‬‬ ‫‪-0.4‬‬ ‫‪-0.5‬‬ ‫‪20‬‬

‫‪0‬‬

‫‪10‬‬

‫‪Lag‬‬

‫‪٣٤‬‬

‫‪ACF‬‬

‫‪-0.2‬‬

‫‪ -٢‬ﻋﻨﺪﻣﺎ ‪θ1 = −0.8‬‬ ‫‪ACF‬‬ ‫‪0.5‬‬ ‫‪0.4‬‬

‫‪0.2‬‬

‫‪ACF‬‬

‫‪0.3‬‬

‫‪0.1‬‬ ‫‪0.0‬‬

‫‪20‬‬

‫‪10‬‬

‫‪0‬‬

‫‪Lag‬‬

‫اﻵن ﻧﺸﺘﻖ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺠﺰﺋﻲ ﻟﻨﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ اﻷوﻟﻲ )‪MA (1‬‬ ‫‪φ00 = 1, by definition‬‬

‫‪φ11 = ρ1 , by definition‬‬ ‫‪1 ρ1‬‬ ‫‪1 ρ1‬‬ ‫) ‪−θ12 (1 − θ12‬‬ ‫‪ρ1 ρ 2‬‬ ‫‪ρ1 0‬‬ ‫‪− ρ12‬‬ ‫‪−θ12‬‬ ‫= ‪φ22‬‬ ‫=‬ ‫=‬ ‫=‬ ‫=‬ ‫‪1 ρ1‬‬ ‫‪1 − ρ12‬‬ ‫‪1 − ρ12 1 + θ12 + θ14‬‬ ‫‪1 − θ16‬‬ ‫‪ρ1 1‬‬ ‫‪ρ1‬‬ ‫‪1 ρ1 ρ1‬‬ ‫‪ρ2‬‬ ‫‪ρ1 1 0‬‬ ‫) ‪−θ13 (1 − θ12‬‬ ‫‪ρ3‬‬ ‫‪0 ρ1 0‬‬ ‫‪ρ13‬‬ ‫=‬ ‫=‬ ‫=‬ ‫‪ρ2‬‬ ‫‪1 ρ1 0 1 − 2 ρ12‬‬ ‫‪1 − θ18‬‬ ‫‪ρ1‬‬ ‫‪ρ1 1 ρ1‬‬ ‫‪1‬‬ ‫‪0 ρ1 1‬‬ ‫وﺑﺸﻜﻞ ﻋﺎم‬ ‫‪k >0‬‬

‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‪:‬‬ ‫‪ -١‬ﻋﻨﺪﻣﺎ ﻋﻨﺪﻣﺎ ‪θ1 = −0.8‬‬

‫‪٣٥‬‬

‫‪ρ1‬‬

‫‪1‬‬

‫‪1‬‬

‫‪ρ1‬‬ ‫‪ρ2‬‬

‫‪ρ1‬‬ ‫‪ρ1‬‬ ‫‪1‬‬

‫‪ρ1‬‬

‫‪(1 − θ ) ,‬‬ ‫‪2‬‬ ‫‪1‬‬

‫)‪2 k +1‬‬

‫‪1‬‬

‫= ‪φ33‬‬

‫‪ρ1‬‬ ‫‪ρ2‬‬ ‫‪k‬‬ ‫‪1‬‬

‫‪−θ‬‬

‫( ‪1 − θ1‬‬

‫= ‪φkk‬‬

‫‪PACF‬‬ ‫‪0.5‬‬ ‫‪0.4‬‬ ‫‪0.3‬‬

‫‪0.1‬‬ ‫‪0.0‬‬

‫‪PACF‬‬

‫‪0.2‬‬

‫‪-0.1‬‬ ‫‪-0.2‬‬ ‫‪-0.3‬‬ ‫‪20‬‬

‫‪0‬‬

‫‪10‬‬

‫‪Lag‬‬

‫‪ -٢‬ﻋﻨﺪﻣﺎ ‪θ1 = 0.8‬‬ ‫‪PACF‬‬ ‫‪0.0‬‬ ‫‪-0.1‬‬

‫‪-0.3‬‬

‫‪PACF‬‬

‫‪-0.2‬‬

‫‪-0.4‬‬ ‫‪-0.5‬‬ ‫‪20‬‬

‫‪0‬‬

‫‪10‬‬

‫‪Lag‬‬

‫ﺧﺎﻣﺴﺎ‪ :‬ﻧﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ اﻟﺜﺎﻧﻴﺔ )‪: ARMA(٠،٢) = MA(٢‬‬ ‫وﺗﻜﺘﺐ ﻋﻠﻲ اﻟﺸﻜﻞ‪:‬‬ ‫‪φ 0 ( B ) zt = δ + θ 2 ( B ) a t‬‬ ‫‪zt = δ + (1 − θ1 B − θ 2 B 2 ) at‬‬

‫) ‪WN ( 0,σ 2‬‬

‫‪zt = δ + at − θ1at −1 − θ 2 at −2 , at‬‬

‫اﻵن ﻧﻮﺟﺪ اﻟﻤﺘﻮﺳﻂ وداﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‪:‬‬ ‫‪E ( zt ) = E (δ + at − θ1at −1 − θ 2 at −2 ) = δ‬‬ ‫‪∴µ = δ‬‬

‫‪٣٦‬‬

‫وﻧﻜﺘﺐ اﻟﻨﻤﻮذج‬ ‫ﺑﻀﺮب هﺬﻩ اﻟﻤﻌﺎدﻟﺔ ﻓﻲ ‪ zt −k − µ‬وأﺧﺬ اﻟﺘﻮﻗﻊ ﻧﺠﺪ‬

‫‪zt − µ = at − θ1at −1 − θ 2 at −2‬‬

‫⎦⎤ ‪E ⎡⎣( zt − µ )( zt −k − µ ) ⎤⎦ = E ⎡⎣ ( zt −k − µ ) at ⎤⎦ − θ1 E ⎡⎣ ( zt −k − µ ) at −1‬‬ ‫‪− θ 2 E ⎡⎣( zt −k − µ ) at −2 ⎤⎦ , k = 0, ±1, ±2,...‬‬

‫او‬ ‫‪− µ ) at −2 ⎤⎦ , k = 0, ±1, ±2,...‬‬

‫‪γ k = E ⎡⎣( zt −k − µ ) at ⎤⎦ − θ1E ⎡⎣( zt −k − µ ) at −1 ⎤⎦ − θ 2 E ⎡⎣( zt −k‬‬

‫وﺑﺤﻠﻬﺎ ﺗﻜﺮارﻳﺎ ﻧﺠﺪ‬ ‫‪2‬‬

‫‪γ 0 = (1 + θ + θ ) σ‬‬ ‫‪2‬‬ ‫‪2‬‬

‫‪2‬‬ ‫‪1‬‬

‫‪γ 1 = ( −θ1 + θ1θ 2 ) σ 2‬‬ ‫‪γ 2 = −θ 2σ 2‬‬ ‫‪γ k = 0, k > 2‬‬

‫وﺑﺎﻟﻘﺴﻤﺔ ﻋﻠﻲ ‪ γ 0‬ﻧﺠﺪ‬

‫‪−θ1 + θ1θ 2‬‬ ‫‪1 + θ12 + θ 22‬‬

‫= ‪ρ1‬‬

‫‪−θ 2‬‬ ‫‪1 + θ12 + θ 22‬‬

‫= ‪ρ2‬‬

‫‪ρ k = 0, k > 2‬‬ ‫وﺗﻜﺘﺐ ﻋﻠﻲ ﺷﻜﻞ داﻟﺔ‬ ‫‪1,‬‬ ‫‪k =0‬‬ ‫⎧‬ ‫‪⎪ −θ + θ θ‬‬ ‫‪⎪ 1 2 1 22 , k = 1‬‬ ‫‪⎪ 1 + θ1 + θ 2‬‬ ‫⎨ = ‪ρk‬‬ ‫‪⎪ −θ 2‬‬ ‫‪, k =2‬‬ ‫‪⎪1 + θ12 + θ 22‬‬ ‫⎪‬ ‫‪0,‬‬ ‫‪k>2‬‬ ‫⎩‬

‫اﻷﺷﻜﺎل اﻟﺘﺎﻟﻴﺔ هﻲ ﻟﺪوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻌﻤﻠﻴﺔ ) ‪MA ( 2‬‬ ‫‪ -٧‬اﻟﺸﻜﻞ )‪θ1 = 0.4, θ 2 = 0.4 (٧‬‬ ‫‪ -٨‬اﻟﺸﻜﻞ )‪θ1 = 1.5, θ 2 = −0.8 (٨‬‬ ‫‪ -٩‬اﻟﺸﻜﻞ )‪θ1 = 0.5, θ 2 = −0.6 (٩‬‬

‫‪٣٧‬‬

(٧) ‫ﺷﻜﻞ‬

ACF 0.0

ACF

-0.1

-0.2

-0.3

0

10

20

Lag

(٨) ‫ﺷﻜﻞ‬

ACF 0.2 0.1 0.0

-0.2 -0.3 -0.4 -0.5 -0.6 -0.7 0

10

20

Lag

(٩) ‫ﺷﻜﻞ‬ ACF 0.4 0.3

0.2 0.1

ACF

ACF

-0.1

0.0 -0.1 -0.2 -0.3 -0.4 -0.5 0

10

Lag

٣٨

20

‫ﻣﻦ اﻟﺼﻌﺐ ﺟﺪا إﻳﺠﺎد ﺷﻜﻞ ﻣﻐﻠﻖ ﻟﺪاﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻨﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ‬ ‫اﻟﺪرﺟﺔ اﻟﺜﺎﻧﻴﺔ ) ‪ MA ( 2‬وﻟﻬﺬا ﺳﻮف ﻧﺴﺘﺨﺪم ﺗﻌﺮﻳﻒ ‪ ١١‬ب ﻟﺤﺴﺎﺑﻬﺎ ورﺳﻤﻬﺎ ﺗﻜﺮارﻳﺎ ﻟﻘﻴﻢ‬ ‫اﻟﻤﻌﺎﻟﻢ اﻟﺘﺎﻟﻴﺔ‪:‬‬ ‫‪ -١٠‬اﻟﺸﻜﻞ )‪θ1 = 0.4, θ 2 = 0.4 (١٠‬‬ ‫‪ -١١‬اﻟﺸﻜﻞ )‪θ1 = 1.5, θ 2 = −0.8 (١١‬‬ ‫‪ -١٢‬اﻟﺸﻜﻞ )‪θ1 = 0.5, θ 2 = −0.6 (١٢‬‬

‫ﺷﻜﻞ )‪(١٠‬‬

‫‪PACF‬‬ ‫‪0.0‬‬

‫‪-0.1‬‬

‫‪PACF‬‬

‫‪-0.2‬‬

‫‪-0.3‬‬

‫‪20‬‬

‫‪0‬‬

‫‪10‬‬

‫‪Lag‬‬

‫ﺷﻜﻞ )‪(١١‬‬ ‫‪PACF‬‬ ‫‪0.2‬‬ ‫‪0.1‬‬ ‫‪0.0‬‬ ‫‪-0.1‬‬

‫‪-0.3‬‬ ‫‪-0.4‬‬ ‫‪-0.5‬‬ ‫‪-0.6‬‬ ‫‪-0.7‬‬

‫‪20‬‬

‫‪10‬‬

‫‪Lag‬‬

‫‪٣٩‬‬

‫‪0‬‬

‫‪PACF‬‬

‫‪-0.2‬‬

‫ﺷﻜﻞ )‪(١٢‬‬ ‫‪PACF‬‬ ‫‪0.4‬‬ ‫‪0.3‬‬ ‫‪0.2‬‬ ‫‪0.1‬‬

‫‪PACF‬‬

‫‪0.0‬‬ ‫‪-0.1‬‬ ‫‪-0.2‬‬ ‫‪-0.3‬‬ ‫‪-0.4‬‬ ‫‪-0.5‬‬ ‫‪20‬‬

‫‪0‬‬

‫‪10‬‬

‫‪Lag‬‬

‫ﺳﺎدﺳﺎ‪ :‬ﻧﻤﻮذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك‪-‬اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ﻣﻦ اﻟﺪرﺟﺔ )‪: ARMA(١،١‬‬ ‫وﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‪:‬‬

‫‪φ1 ( B ) zt = δ + θ1 ( B ) at‬‬

‫‪(1 − φ1B ) zt = δ + (1 − θ1B ) at‬‬ ‫‪WN ( 0, σ 2 ) , φ1 ≠ θ1‬‬

‫‪zt − φ1 zt −1 = δ + at − θ1at −1‬‬ ‫‪zt = δ + φ1 zt −1 + at − θ1at −1 , at‬‬

‫ﺷﺮط اﻹﺳﺘﻘﺮار ‪ φ1 < 1‬وﺷﺮط اﻹﻧﻘﻼب ‪ θ1 < 1‬وهﻨﺎك ﺷﺮط ﺁﺧﺮ ﻳﺴﻤﻰ ﺷﺮط اﻹﻣﺘﺴﺎخ‬ ‫‪ Degeneracy Condition‬وهﻮ ‪ φ1 ≠ θ1‬وهﺬا اﻟﺸﺮط ﻳﻀﻤﻦ ﻋﺪم إﻣﺘﺴﺎخ اﻟﻨﻤﻮذج إﻟﻰ ﻧﻤﻮذج‬ ‫أﻗﻞ درﺟﺔ ﻓﻔﻲ ﺣﺎﻟﺔ آﻮن ‪ φ1 = θ1‬ﻓﻤﻦ اﻟﻌﻼﻗﺔ ‪ (1 − φ1B ) zt = δ + (1 − θ1B ) at‬وﺑﺎﻟﻘﺴﻤﺔ ﻋﻠﻰ‬ ‫) ‪ (1 − φ1B‬ﻧﺠﺪ أن اﻟﻨﻤﻮذج ﻳﺼﺒﺢ ‪ zt = δ ′ + at‬ﺣﻴﺚ‬ ‫ﻧﻮﺟﺪ اﻟﻤﺘﻮﺳﻂ آﺎﻟﺘﺎﻟﻲ‪:‬‬

‫‪δ‬‬ ‫‪1 − φ1‬‬

‫= ‪ δ ′‬وهﻮ ) ‪ARMA ( 0,0‬‬

‫‪(1 − φ1B ) zt = δ + (1 − θ1B ) at‬‬ ‫‪(1 − θ1B ) a‬‬ ‫‪δ‬‬ ‫= ‪zt‬‬ ‫‪+‬‬ ‫‪t‬‬ ‫) ‪1 − φ1 (1 − φ1B‬‬ ‫‪(1 − θ1B ) E a‬‬ ‫‪δ‬‬ ‫= ) ‪E ( zt‬‬ ‫‪+‬‬ ‫)‪( t‬‬ ‫) ‪1 − φ1 (1 − φ1B‬‬

‫وذﻟﻚ ﻷن ‪ φ1 < 1‬وهﻜﺬا‬ ‫‪δ‬‬

‫أي‬

‫‪δ‬‬ ‫‪1 − φ1‬‬

‫‪1 − φ1‬‬

‫= ) ‪E ( zt‬‬

‫= ‪ E ( zt ) = µ‬أو ) ‪ δ = µ (1 − φ1‬وﺑﺎﻟﺘﻌﻮﻳﺾ ﻋﻦ ‪ δ‬ﻧﺠﺪ‬ ‫‪zt = µ (1 − φ1 ) + φ1 zt −1 + at − θ1at −1‬‬

‫‪( zt − µ ) − φ1 ( zt −1 − µ ) = at − θ1at −1‬‬ ‫وﺑﻀﺮب ﻃﺮﻓﻲ اﻟﻤﻌﺎدﻟﺔ ﺑﺎﻟﺤﺪ ‪ ( zt −k − µ ) , k = 0, ±1, ±2,...‬وأﺧﺬ اﻟﺘﻮﻗﻊ ﻟﻠﻄﺮﻓﻴﻦ ﻧﺠﺪ‬ ‫‪E ⎡⎣( zt −k − µ )( zt − µ )⎤⎦ − φ1E ⎡⎣ ( zt −k − µ )( zt −1 − µ )⎤⎦ = E ⎡⎣( zt −k − µ ) at ⎤⎦ − θ1E ⎡⎣( zt −k − µ ) at −1 ⎤⎦ ,‬‬ ‫‪k = 0, ±1, ±2,...‬‬

‫وﻣﻨﻬﺎ‬ ‫‪٤٠‬‬

γ k − φ1γ k −1 = E ⎡⎣( zt −k − µ ) at ⎤⎦ − θ1E ⎡⎣ ( zt −k − µ ) at −1 ⎤⎦ , k = 0, ±1, ±2,...

‫وﺑﺤﻠﻬﺎ ﺗﻜﺮارﻳﺎ ﻧﺠﺪ‬

k = 0 γ 0 − φ1γ 1 = E ⎡⎣ ( zt − µ ) at ⎤⎦ − θ1E ⎡⎣ ( zt − µ ) at−1 ⎤⎦

‫ ﺑﻀﺮب اﻟﻌﻼﻗﺔ‬E ⎡⎣( zt − µ ) at −1 ⎤⎦ ‫ و‬E ⎡⎣( zt − µ ) at ⎤⎦ ‫ﻧﻮﺟﺪ اﻵن آﻞ ﻣﻦ‬

( zt − µ ) − φ1 ( zt −1 − µ ) = at − θ1at −1

‫ وأﺧﺬ اﻟﺘﻮﻗﻊ‬at −1 ‫ و‬at ‫ﻓﻲ آﻞ ﻣﻦ‬ E ⎡⎣ ( zt − µ ) at ⎤⎦ − φ1E ⎡⎣ ( zt −1 − µ ) at ⎤⎦ = E [at at ] − θ1 E [at −1at ]

‫ ﻧﺠﺪ‬١ ‫وﻣﻦ اﻟﻘﺎﻋﺪة‬

E ⎡⎣( zt − µ ) at ⎤⎦ − φ1 ( 0 ) = σ − θ1 ( 0 ) 2

E ⎡⎣( zt − µ ) at ⎤⎦ = σ 2 E ⎡⎣( zt − µ ) at −1 ⎤⎦ − φ1E ⎡⎣( zt −1 − µ ) at −1 ⎤⎦ = E [at at −1 ] − θ1E [at −1at −1 ]

‫و‬

E ⎡⎣( zt − µ ) at −1 ⎤⎦ − φ1σ 2 = 0 − θ1σ 2 ∴ E ⎡⎣( zt − µ ) at −1 ⎤⎦ = σ 2 (φ1 − θ1 )

‫وﺑﺎﻟﺘﻌﻮﻳﺾ ﻓﻲ اﻟﺼﻴﻐﺔ اﻟﺴﺎﺑﻘﺔ ﻧﺠﺪ‬ k = 0 γ 0 − φ1γ 1 = σ 2 − θ1σ 2 (φ1 − θ1 ) ∴γ 0 − φ1γ 1 = σ 2 ⎡⎣1 − θ1 (φ1 − θ1 )⎤⎦ k = 1 γ 1 − φ1γ 0 = E ⎡⎣ ( zt −1 − µ ) at ⎤⎦ − θ1 E ⎡⎣ ( zt −1 − µ ) at −1 ⎤⎦

‫و‬

∴ γ 1 − φ1γ 0 = −θ1σ 2 k = 2 γ 2 − φ1γ 1 = E ⎡⎣ ( zt −2 − µ ) at ⎤⎦ − θ1 E ⎡⎣ ( zt −2 − µ ) at −1 ⎤⎦ = 0

‫و‬

∴ k ≥ 2 γ k − φ1γ k −1 = 0

‫وﻣﻦ اﻟﻤﻌﺎدﻻت‬

γ 0 − φ1γ 1 = σ ⎡⎣1 − θ1 (φ1 − θ1 )⎤⎦ 2

‫و‬ γ 1 − φ1γ 0 = −θ1σ

2

‫ﻧﺠﺪ‬ γ0 = γ1 =

ρ1 =

1 + θ − 2φ1θ1 2 σ 1 − φ12 2 1

(1 − φ1θ1 )(φ1 − θ1 ) σ 2 1 − φ12

‫وﻣﻦ اﻟﻌﻼﻗﺘﻴﻦ اﻟﺴﺎﺑﻘﺘﻴﻦ ﻧﺠﺪ‬

γ 1 (1 − φ1θ1 )(φ1 − θ1 ) = 1 + θ12 − 2φ1θ1 γ0

٤١

‫وﻣﻦ اﻟﻌﻼﻗﺔ‬

‫‪γ k − φ1γ k −1 = 0, k ≥ 2‬‬

‫وﺑﺎﻟﻘﺴﻤﺔ ﻋﻠﻰ ‪ γ 0‬ﻧﺠﺪ‬ ‫‪ρ k − φ1 ρ k −1 = 0, k ≥ 2‬‬ ‫وﻳﻤﻜﻦ ﺣﻞ هﺬﻩ اﻟﻤﻌﺎدﻟﺔ ﺗﻜﺮارﻳﺎ ﻟﺠﻤﻴﻊ ﻗﻴﻢ ‪ k ≥ 2‬ﺑﺈﺳﺘﺨﺪام اﻟﻘﻴﻢ اﻷوﻟﻴﺔ ‪ ρ0 = 1‬و‬ ‫) ‪(1 − φ1θ1 )(φ1 − θ1‬‬ ‫= ‪ ρ1‬ﻓﻤﺜﻼ‬ ‫‪1 + θ12 − 2φ1θ1‬‬ ‫‪ρ2 = φ1 ρ1‬‬ ‫) ‪(1 − φ1θ1 )(φ1 − θ1‬‬ ‫‪ρ 2 = φ1‬‬ ‫‪1 + θ12 − 2φ1θ1‬‬ ‫‪ρ3 = φ1 ρ 2‬‬ ‫) ‪(1 − φ1θ1 )(φ1 − θ1‬‬ ‫‪ρ3 = φ12‬‬ ‫‪1 + θ12 − 2φ1θ1‬‬

‫وهﻜﺬا‪.‬‬ ‫ﻧﻜﺘﺐ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻨﻤﻮذج )‪ ARMA (1,1‬ﻋﻠﻰ اﻟﺸﻜﻞ‬ ‫‪1,‬‬ ‫‪k =0‬‬ ‫⎧‬ ‫⎪‬ ‫) ‪⎪ (1 − φ1θ1 )(φ1 − θ1‬‬ ‫‪, k =1‬‬ ‫⎨ = ‪ρk‬‬ ‫‪2‬‬ ‫‪⎪ 1 + θ1 − 2φ1θ1‬‬ ‫⎩⎪‬ ‫‪k≥2‬‬ ‫‪φ1 ρ k −1‬‬

‫ﺷﻜﻞ ‪ ١٣‬ﻳﻌﻄﻲ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻘﻴﻢ ‪φ1 = 0.9,θ1 = −0.5‬‬

‫ﺷﻜﻞ)‪(١٣‬‬

‫) ‪A C F o f A R M A (1 ,1‬‬ ‫‪1 .0‬‬ ‫‪0 .9‬‬ ‫‪0 .8‬‬ ‫‪0 .7‬‬ ‫‪0 .6‬‬ ‫‪0 .4‬‬ ‫‪0 .3‬‬ ‫‪0 .2‬‬ ‫‪0 .1‬‬ ‫‪0 .0‬‬

‫‪15‬‬

‫‪5‬‬

‫‪10‬‬

‫‪Lag‬‬

‫ﺷﻜﻞ ‪ ١٤‬ﻳﻌﻄﻲ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻘﻴﻢ ‪φ1 = −0.9,θ1 = −0.5‬‬

‫‪٤٢‬‬

‫‪0‬‬

‫‪C1‬‬

‫‪0 .5‬‬

‫ﺷﻜﻞ)‪(١٤‬‬ ‫) ‪A C F o f A R M A (1 ,1‬‬ ‫‪0 .5‬‬

‫‪C1‬‬

‫‪0 .0‬‬

‫‪-0 .5‬‬

‫‪10‬‬

‫‪15‬‬

‫‪0‬‬

‫‪5‬‬

‫‪Lag‬‬

‫ﻧﻼﺣﻆ ان داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻨﻤﻮذج )‪ ARMA (1,1‬ﺗﺘﺨﺎﻣﺪ اﺳﻴﺎ ﻓﻲ إﺗﺠﺎﻩ واﺣﺪ أو ﻣﺘﺮدد ﺑﻴﻦ‬ ‫اﻟﻘﻴﻢ اﻟﻤﻮﺟﺒﺔ واﻟﺴﺎﻟﺒﺔ وهﻲ ﻓﻲ هﺬا ﺗﺸﺒﻪ ﺗﻤﺎﻣﺎ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻨﻤﻮذج )‪ AR (1‬ﻣﺎﻋﺪى ان‬ ‫اﻟﺘﺨﺎﻣﺪ ﻳﺒﺪأ ﻣﻦ ‪ ) ρ1‬ﺑﺮهﻦ أن ‪( ρ k = φ1k −1 ρ1 , k ≥ 2‬‬ ‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ‪ φkk‬ﺗﺤﺴﺐ ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ١١‬أو ‪١١‬ب آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫ﻣﻦ ﺗﻌﺮﻳﻒ ‪١١‬ب ﻧﻮﺟﺪ ‪ φkk‬ﺗﻜﺮارﻳﺎ‬ ‫‪φ00 = 1, by definition‬‬

‫) ‪(1 − φ1θ1 )(φ1 − θ1‬‬ ‫‪1 + θ12 − 2φ1θ1‬‬

‫= ‪φ11 = ρ1‬‬

‫‪ρ 2 − φ11 ρ1‬‬ ‫‪1 − φ11 ρ1‬‬ ‫‪ρ −φ ρ −φ ρ‬‬ ‫‪φ33 = 3 21 2 22 1 , φ21 = φ11 − φ22φ11‬‬ ‫‪1 − φ21 ρ1 − φ22 ρ 2‬‬

‫= ‪φ22‬‬

‫وهﻜﺬا ﺗﺤﺴﺐ ﺑﻘﻴﺔ اﻟﻘﻴﻢ ﺗﻜﺮارﻳﺎ‪.‬‬ ‫ﻓﻤﺜﻼ ﻟﻠﻘﻴﻢ ‪ φ1 = 0.9,θ1 = −0.5‬ﻧﺠﺪ‬

‫‪φ11 = 0.944186 φ22 = -0.384471 φ33 = 0.183710‬‬ ‫‪φ44 = -0.908462 φ55 = 0.452979 φ66 = -0.226337‬‬ ‫‪φ77 = 0.113154 φ88 = -0.565702 φ99 = 0.282834‬‬

‫وﻧﺮﺳﻢ هﺬﻩ اﻟﻘﻴﻢ ﻓﻲ ﺷﻜﻞ ‪١٥‬‬ ‫ﺷﻜﻞ ‪١٥‬‬ ‫) ‪P A C F o f A R M A (1 ,1‬‬ ‫‪1 .0‬‬

‫‪0 .5‬‬

‫‪C2‬‬ ‫‪0 .0‬‬

‫‪15‬‬

‫‪5‬‬

‫‪10‬‬

‫‪Lag‬‬

‫‪٤٣‬‬

‫‪0‬‬

‫ﺷﻜﻞ ‪ ١٦‬ﻳﺒﻴﻦ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻘﻴﻢ ‪φ1 = −0.9,θ1 = −0.5‬‬

‫ﺷﻜﻞ ‪١٦‬‬ ‫) ‪P A C F o f A R M A (1 ,1‬‬ ‫‪0 .3‬‬ ‫‪0 .2‬‬ ‫‪0 .1‬‬ ‫‪0 .0‬‬

‫‪-0 .2‬‬

‫‪C2‬‬

‫‪-0 .1‬‬

‫‪-0 .3‬‬ ‫‪-0 .4‬‬ ‫‪-0 .5‬‬ ‫‪-0 .6‬‬

‫‪15‬‬

‫‪5‬‬

‫‪10‬‬

‫‪0‬‬

‫‪Lag‬‬

‫ﻧﻼﺣﻆ ان داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻨﻤﻮذج )‪ ARMA (1,1‬ﺗﺘﺨﺎﻣﺪ اﺳﻴﺎ ﻓﻲ إﺗﺠﺎﻩ واﺣﺪ أو ﻣﺘﺮدد‬ ‫ﺑﻴﻦ اﻟﻘﻴﻢ اﻟﻤﻮﺟﺒﺔ واﻟﺴﺎﻟﺒﺔ وهﻲ ﻓﻲ هﺬا ﺗﺸﺒﻪ ﺗﻤﺎﻣﺎ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻨﻤﻮذج )‪MA (1‬‬ ‫ﻣﺎﻋﺪى ان اﻟﺘﺨﺎﻣﺪ ﻳﺒﺪأ ﺑﻌﺪ اﻟﻘﻴﻤﺔ اﻷوﻟﻴﺔ ‪. φ11 = ρ1‬‬

‫ﺧﻮاص ﻧﻤﺎذج )‪: ARMA(p,q‬‬ ‫أوﻻ‪ :‬ﻧﻤﻮذج )‪AR(p‬‬ ‫وﻳﺘﻤﻴﺰ ﺑﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫‪ -١‬داﻟﺔ ﺗﺮاﺑﻂ ذاﺗﻲ ﺗﻤﺘﺪ ﻻﻧﻬﺎﺋﻴﺎ وﺗﺘﻜﻮن ﻣﻦ ﺧﻠﻴﻂ ﻣﻦ اﻟﺘﺨﺎﻣﺪات اﻻﺳﻴﺔ واﻟﺘﺨﺎﻣﺪات اﻟﺠﻴﺒﻴﺔ‪.‬‬ ‫‪ -٢‬داﻟﺔ ﺗﺮاﺑﻂ ذاﺗﻲ ﺟﺰﺋﻲ ﺗﺘﻜﻮن ﻣﻦ أﺻﻔﺎر ﻟﻘﻴﻢ اﻟﺘﺨﻠﻔﺎت ‪ k > p‬أي‬ ‫‪φ11 = φ22 = φ33 = L = φ pp ≠ 0‬‬ ‫‪φ p +1, p +1 = φ p +2, p + 2 = L = 0‬‬

‫وﻳﺴﻤﻰ هﺬا ﻗﻄﻌﺎ ﻓﻲ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﺑﻌﺪ اﻟﺘﺨﻠﻒ ‪. k > p‬‬ ‫ﺛﺎﻧﻴﺎ‪ :‬ﻧﻤﻮذج )‪: MA(q‬‬ ‫وﻳﺘﻤﻴﺰ ﺑﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫‪ -١‬داﻟﺔ ﺗﺮاﺑﻂ ذاﺗﻲ ﺗﺘﻜﻮن ﻣﻦ أﺻﻔﺎر ﻟﻘﻴﻢ اﻟﺘﺨﻠﻔﺎت ‪ k > q‬أي‬ ‫‪ρ1 = ρ 2 = ρ3 = L = ρ q ≠ 0‬‬ ‫‪ρ q+1,q+1 = ρ q+2,q +2 = L = 0‬‬

‫وﻳﺴﻤﻰ هﺬا ﻗﻄﻌﺎ ﻓﻲ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﺑﻌﺪ اﻟﺘﺨﻠﻒ ‪. k > q‬‬ ‫‪ -٢‬داﻟﺔ ﺗﺮاﺑﻂ ذاﺗﻲ ﺟﺰﺋﻲ ﺗﻤﺘﺪ ﻻﻧﻬﺎﺋﻴﺎ وﺗﺘﻜﻮن ﻣﻦ ﺧﻠﻴﻂ ﻣﻦ اﻟﺘﺨﺎﻣﺪات اﻻﺳﻴﺔ واﻟﺘﺨﺎﻣﺪات‬ ‫اﻟﺠﻴﺒﻴﺔ‪.‬‬ ‫ﻻﺣﻆ اﻹزدواﺟﻴﺔ ‪ Duality‬ﺑﻴﻦ ﻧﻤﻮذﺟﻲ ‪ AR‬و ‪.MA‬‬ ‫ﺛﺎﻟﺜﺎ‪ :‬اﻟﻨﻤﻮذج اﻟﻤﺨﺘﻠﻂ )‪:ARMA(p,q‬‬ ‫وﻳﺘﻤﻴﺰ ﺑﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻨﻤﻮذج اﻟﻤﺨﺘﻠﻂ ﺗﻤﺘﺪ ﻻﻧﻬﺎﺋﻴﺎ وﺗﺘﻜﻮن ﻣﻦ ﺧﻠﻴﻂ ﻣﻦ‬ ‫اﻟﺘﺨﺎﻣﺪات اﻻﺳﻴﺔ واﻟﺘﺨﺎﻣﺪات اﻟﺠﻴﺒﻴﺔ اﻟﺘﻲ ﺗﻨﺘﻬﻲ إﻟﻰ اﻟﺼﻔﺮ آﻠﻤﺎ زاد اﻟﺘﺨﻠﻒ ‪ . k‬ﻋﻨﺪﻣﺎ ﺗﻜﻮن‬ ‫‪ k > q − p‬ﻓﺈن داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﺗﺘﺤﺪد ﻣﻦ ﺟﺰء اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ﻟﻠﻨﻤﻮذج و ﻋﻨﺪﻣﺎ ﺗﻜﻮن‬ ‫‪ k > p − q‬ﻓﺈن داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﺗﺘﺤﺪد ﻣﻦ ﺟﺰء اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻟﻠﻨﻤﻮذج‪.‬‬ ‫‪٤٤‬‬

‫اﻟﻔﺼﻞ اﻟﺜﺎﻟﺚ‬ ‫ﻧﻤﺎذج اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ ﻏﻴﺮ اﻟﻤﺴﺘﻘﺮة ‪Nonstationar Time‬‬ ‫‪:Series Models‬‬ ‫اوﻻ‪ :‬ﻋﺪم اﻹﺳﺘﻘﺮار ﻓﻲ اﻟﻤﺘﻮﺳﻂ‪:‬‬ ‫ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ٦‬ﻹﺳﺘﻘﺮار ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻧﺮى ان اﻟﺸﺮط اﻷول ﻟﻺﺳﺘﻘﺮار‬ ‫‪ E ( zt ) = µ = constant ∀t‬ﻳﺘﻄﻠﺐ أن ﻳﻜﻮن ﻣﺘﻮﺳﻂ اﻟﻤﺘﺴﻠﺴﻠﺔ ﺛﺎﺑﺖ ﻋﻠﻰ ﻃﻮل اﻟﺰﻣﻦ‪ ،‬ﻓﻤﺜﻼ‬ ‫ﻟﻨﻤﻮذج اﻹﻧﺠﺮاف اﻟﺨﻄﻲ‬ ‫) ∞ ‪WN ( 0, σ 2 ) , b0 , b1 ∈ ( −∞,‬‬

‫‪zt = b0 + b1t + at , at‬‬

‫ﻧﺠﺪ ان اﻟﻤﺘﻮﺳﻂ هﻮ‬

‫‪E ( z )t = b0 + b1t‬‬

‫وهﻮ ﻏﻴﺮ ﺛﺎﺑﺖ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﺰﻣﻦ‪ ،‬اي ان ﺷﺮط اﻹﺳﺘﻘﺮار اﻷول ﻏﻴﺮ ﻣﺘﺤﻘﻖ ﻓﻲ هﺬﻩ اﻟﺤﺎﻟﺔ‪.‬‬ ‫ﻟﻨﺤﺎول اﻟﺘﺤﻮﻳﻞ ‪ ∇zt‬وذﻟﻚ ﺑﺘﻄﺒﻴﻖ ﻋﺎﻣﻞ اﻟﺘﻔﺮﻳﻖ ﻋﻠﻰ اﻟﻨﻤﻮذج ﻧﺠﺪ‬

‫‪wt = ∇zt = zt − zt −1 = b0 + b1t + at − b0 − b1 ( t − 1) − at −1‬‬ ‫‪=b1 + at − at −1 = b1 + ct‬‬

‫) ‪WN ( 0,ν 2‬‬

‫) ﺗﻤﺮﻳﻦ‪ :‬أوﺟﺪ اﻟﻌﻼﻗﺔ ﺑﻴﻦ ‪ ν 2‬و ‪( σ 2‬‬ ‫اﻵن ﻧﺠﺪ ﻣﺘﻮﺳﻂ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺠﺪﻳﺪة ‪wt‬‬

‫‪∴ wt = b1 + ct , ct‬‬

‫‪E ( wt ) = b1 = constant ∀t‬‬

‫أي ان ﺗﻄﺒﻴﻖ اﻟﺘﺤﻮﻳﻞ ) ‪ ∇ = (1 − B‬ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻏﻴﺮ اﻟﻤﺴﺘﻘﺮة ) أي أﺧﺬ اﻟﺘﻔﺮﻳﻖ اﻷول‬ ‫ﻟﻠﻤﺘﺴﻠﺴﻠﺔ( ﺣﻮﻟﻬﺎ إﻟﻰ ﻣﺘﺴﻠﺴﻠﺔ ﻣﺴﺘﻘﺮة‪.‬‬ ‫آﻤﺜﺎل ﺁﺧﺮ ﻧﻤﻮذج اﻹﻧﺠﺮاف اﻟﺘﺮﺑﻴﻌﻲ‬ ‫) ∞ ‪WN ( 0, σ 2 ) , b0 , b1 , b2 ∈ ( −∞,‬‬

‫‪zt = b0 + b1t + b2t 2 + at , at‬‬

‫ﺑﺈﻳﺠﺎد اﻟﻤﺘﻮﺳﻂ‬ ‫‪2‬‬

‫‪E ( zt ) = b0 + b1t + b2t‬‬

‫وهﻮ ﻳﻌﺘﻤﺪ ﻋﻠﻰ اﻟﺰﻣﻦ‪ ،‬أي ان اﻟﻨﻤﻮذج ﻏﻴﺮ ﻣﺴﺘﻘﺮ‪ .‬ﺑﺄﺧﺬ اﻟﺘﺤﻮﻳﻞ ‪ ) ∇ 2 zt‬أﺧﺬ اﻟﺘﻔﺮﻳﻖ اﻟﺜﺎﻧﻲ(‬ ‫ﻧﺠﺪ‬

‫) ‪∇2 zt = ∇ 2 ( b0 + b1t + b2t 2 + at‬‬

‫) ‪+ b1t + b2t 2 + at‬‬

‫‪(1 − 2 B + B ) z = (1 − 2 B + B )( b‬‬ ‫‪2‬‬

‫‪2‬‬

‫‪0‬‬

‫‪t‬‬

‫‪wt = {b0 − 2b0 + b0 } + {b1t − 2b1 ( t − 1) + b1 ( t − 2 )} +‬‬

‫}‬

‫‪− 2b2 ( t − 1) + b2 ( t − 2 ) +‬‬ ‫‪2‬‬

‫‪2‬‬

‫‪2‬‬

‫‪{b t‬‬ ‫‪2‬‬

‫} ‪{at − 2at −1 + at −2‬‬ ‫} ‪= 2b2 + {at − 2at −1 + at −2‬‬ ‫) ‪WN ( 0,τ 2‬‬

‫وهﻜﺬا‬

‫‪٤٥‬‬

‫‪=b′ + ht , ht‬‬

‫) ‪WN ( 0,τ 2‬‬

‫‪wt = ∇2 zt = b′ + ht , ht‬‬

‫‪E ( wt ) = b′ = constant ∀t‬‬

‫أي ان ﺗﻄﺒﻴﻖ اﻟﺘﺤﻮﻳﻞ ‪) ∇2‬أي اﺧﺬ اﻟﺘﻔﺮﻳﻖ اﻟﺜﺎﻧﻲ( ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻏﻴﺮ اﻟﻤﺴﺘﻘﺮة ﺣﻮﻟﻬﺎ اﻟﻰ‬ ‫ﻣﺴﺘﻘﺮة‪.‬‬ ‫‪2‬‬ ‫‪2‬‬ ‫) ﺗﻤﺮﻳﻦ‪ :‬أوﺟﺪ اﻟﻌﻼﻗﺔ ﺑﻴﻦ ‪ τ‬و ‪.( σ‬‬ ‫ﺑﺸﻜﻞ ﻋﺎم إذا آﺎن اﻟﻨﻤﻮذج ﻏﻴﺮ اﻟﻤﺴﺘﻘﺮ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫) ∞ ‪WN ( 0, σ 2 ) , b0 , b1 ,L , bd ∈ ( −∞,‬‬

‫‪zt = b0 + b1t + L + bd t + at , at‬‬ ‫‪d‬‬

‫ﻓﺈن اﻟﺘﺤﻮﻳﻞ ‪ ∇d zt‬ﻳﺤﻮﻟﻪ إﻟﻰ ﻧﻤﻮذج ﻣﺴﺘﻘﺮ‪ ،‬أي ان ‪ wt = ∇ d zt‬هﻮ ﻧﻤﻮذج ﻣﺴﺘﻘﺮ‪.‬‬

‫ﺗﻌﺮﻳﻒ ‪ :١٦‬ﻟﻠﻤﺘﺴﻠﺴﻠﺔ ﻏﻴﺮ اﻟﻤﺴﺘﻘﺮة ﻓﻲ اﻟﻤﺘﻮﺳﻂ‬ ‫) ∞ ‪zt = b0 + b1t + L + bd t + at , at WN ( 0, σ 2 ) , b0 , b1 ,L , bd ∈ ( −∞,‬‬ ‫اﻟﺘﺤﻮﻳﻞ ‪ ∇d zt‬وهﻮ اﻟﺘﻔﺮﻳﻖ ﻟﻠﺪرﺟﺔ ‪ d‬ﻟﻠﻤﺘﺴﻠﺴﻠﺔ ﻳﺤﻮﻟﻬﺎ إﻟﻰ ﻣﺘﺴﻠﺴﻠﺔ ﻣﺴﺘﻘﺮة‪.‬‬ ‫‪d‬‬

‫ﺛﺎﻧﻴ ًﺎ‪ :‬ﻋﺪم اﻹﺳﺘﻘﺮار ﻓﻲ اﻟﺘﺒﺎﻳﻦ‪:‬‬ ‫ﻣﻦ ﺗﻌﺮﻳﻒ ‪ ٦‬ﻹﺳﺘﻘﺮار ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ‪ ،‬اﻟﺸﺮط اﻟﺜﺎﻧﻲ‬

‫ﻳﺘﻄﻠﺐ أن ﻳﻜﻮن اﻟﺘﺒﺎﻳﻦ ﺛﺎﺑﺖ ﻟﺠﻤﻴﻊ ﻗﻴﻢ ‪. t‬‬ ‫ﻓﻤﺜﻼ ﻟﻨﻤﻮذج اﻟﻤﺸﻲ اﻟﻌﺸﻮاﺋﻲ‬

‫‪V ( zt ) = γ 0 = constant ∀t‬‬

‫) ‪WN ( 0, σ 2‬‬

‫‪zt = zt −1 + at , at‬‬

‫ﻧﺠﺪ ﻣﻦ اﻟﺘﻌﻮﻳﺾ اﻟﻤﺘﻜﺮر‬ ‫‪zt = a1 + a2 + L + at‬‬

‫وﺑﺈﺧﺬ اﻟﺘﻮﻗﻊ واﻟﺘﺒﺎﻳﻦ‬

‫‪E ( zt ) = 0 = constant ∀t‬‬ ‫‪V ( zt ) = tσ 2‬‬

‫وﻧﻼﺣﻆ أن اﻟﺘﺒﺎﻳﻦ ﻳﻌﺘﻤﺪ ﻋﻠﻰ اﻟﺰﻣﻦ ‪. t‬‬ ‫ﺑﺄﺧﺬ اﻟﺘﻔﺮﻳﻖ اﻷول‬ ‫‪wt = ∇zt = zt − zt −1 = at‬‬

‫وﺑﺈﺧﺬ اﻟﺘﻮﻗﻊ واﻟﺘﺒﺎﻳﻦ‬ ‫‪E ( wt ) = 0 = constant ∀t‬‬ ‫‪V ( wt ) = σ 2 = constant ∀t‬‬

‫إذًا اﻟﺘﻔﺮﻳﻖ اﻷول ﺣﻮل اﻟﻤﺘﺴﻠﺴﻠﺔ ﻏﻴﺮ اﻟﻤﺴﺘﻘﺮة ﻓﻲ اﻟﺘﺒﺎﻳﻦ إﻟﻰ ﻣﺘﺴﻠﺴﻠﺔ ﻣﺴﺘﻘﺮة‪.‬‬ ‫ﺑﺸﻜﻞ ﻋﺎم ﻟﻮ آﺎن اﻟﺘﺒﺎﻳﻦ داﻟﺔ ﻟﻤﺴﺘﻮى )ﻣﺘﻮﺳﻂ( ﻣﺘﻐﻴﺮ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫) ‪V ( zt ) = cf ( µt‬‬

‫ﺣﻴﺚ ‪ c > 0‬ﺛﺎﺑﺖ و )⋅( ‪ f‬داﻟﺔ ﻣﻌﺮوﻓﺔ ﺗﻌﻄﻰ ﻗﻴﻤﺔ ﻏﻴﺮ ﺳﺎﻟﺒﺔ و ‪ µt‬ﻣﺴﺘﻮى أو ﻣﺘﻮﺳﻂ ﻳﺘﻐﻴﺮ‬ ‫ﻣﻊ اﻟﺰﻣﻦ و ﺑﺎﻟﺘﺎﻟﻲ ﻓﺈن اﻟﺘﺒﺎﻳﻦ ﻳﻌﺘﻤﺪ ﻋﻠﻰ اﻟﺰﻣﻦ وهﻨﺎ ﻧﺤﺎول إﻳﺠﺎد ﺗﺤﻮﻳﻞ ) ‪ T ( zt‬أي إﻳﺠﺎد داﻟﺔ‬ ‫)⋅ ( ‪ T‬ﻹﺳﺘﻘﺮار اﻟﺘﺒﺎﻳﻦ‪.‬‬ ‫‪٤٦‬‬

‫اﻟﺘﺤﻮﻳﻞ‬ ‫‪λ‬‬

‫‪zt − 1‬‬

‫‪λ‬‬

‫= ) ‪yt = T ( zt‬‬

‫ﻳﻌﻄﻲ ﻣﺘﺴﻠﺴﻠﺔ ﻣﺴﺘﻘﺮة ﻓﻲ اﻟﺘﺒﺎﻳﻦ ﺣﻴﺚ ) ∞ ‪ λ ∈ ( −∞,‬هﻮ ﻣﻌﻠﻢ اﻟﺘﺤﻮﻳﻞ‪ .‬اﻟﺠﺪول اﻟﺘﺎﻟﻲ ﻳﻌﻄﻲ‬ ‫اﻟﻘﻴﻢ اﻷآﺜﺮ إﺳﺘﺨﺪاﻣﺎ ﻟﻠﻤﻌﻠﻢ ‪ λ‬ﻣﻊ اﻟﺘﺤﻮﻳﻼت اﻟﻤﻘﺎﺑﻠﺔ ﻟﻬﺎ‪:‬‬ ‫‪1.0‬‬ ‫‪zt‬‬

‫‪0.5‬‬

‫‪0.0‬‬

‫‪-0.5‬‬

‫‪zt‬‬

‫‪ln zt‬‬

‫‪1‬‬ ‫‪zt‬‬

‫‪-0.1‬‬

‫‪λ‬‬

‫‪1‬‬ ‫‪zt‬‬

‫‪yt‬‬

‫ﻣﺜﺎل‪:‬‬ ‫اﻟﺸﻜﻞ)ا( ﻟﻤﺘﺴﻠﺴﻠﺔ ﻏﻴﺮ ﻣﺴﺘﻘﺮة ﻓﻲ اﻟﻤﺘﻮﺳﻂ واﻟﺘﺒﺎﻳﻦ ‪zt‬‬ ‫‪O r ig in a l S e r ie s‬‬ ‫‪400‬‬

‫‪300‬‬

‫)‪z(t‬‬ ‫‪200‬‬

‫‪100‬‬ ‫‪90‬‬

‫‪80‬‬

‫‪70‬‬

‫‪60‬‬

‫‪50‬‬

‫‪40‬‬

‫‪30‬‬

‫‪20‬‬

‫‪10‬‬

‫‪In d e x‬‬

‫اﻟﺸﻜﻞ)ب( اﻟﻤﺘﺴﻠﺴﻠﺔ ﺑﻌﺪ ﺗﺜﺒﻴﺖ اﻟﺘﺒﺎﻳﻦ ﺑﺈﺟﺮاء اﻟﺘﺤﻮﻳﻞ ‪yt = ln zt‬‬ ‫‪T r a n s f o r m e d S e r ie s‬‬ ‫‪6 .0‬‬

‫‪5 .5‬‬

‫)‪ln z(t‬‬ ‫‪5 .0‬‬

‫‪90‬‬

‫‪80‬‬

‫‪70‬‬

‫‪60‬‬

‫‪50‬‬

‫‪40‬‬

‫‪30‬‬

‫‪20‬‬

‫اﻟﺸﻜﻞ)ج( اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﺤﻮﻟﺔ ‪ yt‬ﺑﻌﺪ إﺟﺮاء اﻟﺘﻔﺮﻳﻖ اﻷول ‪∇yt = yt − yt −1‬‬

‫‪٤٧‬‬

‫‪10‬‬

‫‪In d e x‬‬

‫‪D if f e re n c e d a n d T ra n s f o rm e d S e rie s‬‬ ‫‪0 .2‬‬

‫‪0 .1‬‬ ‫)‪y(t)-y(t-1‬‬

‫‪0 .0‬‬

‫‪-0 .1‬‬

‫‪-0 .2‬‬

‫‪90‬‬

‫‪80‬‬

‫‪70‬‬

‫‪60‬‬

‫‪50‬‬

‫‪40‬‬

‫‪30‬‬

‫‪20‬‬

‫‪10‬‬

‫‪In d e x‬‬

‫ﻻﺣﻆ آﻴﻒ اﺻﺒﺤﺖ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻣﺴﺘﻘﺮة ﻓﻲ آﻞ ﻣﻦ اﻟﻤﺘﻮﺳﻂ واﻟﺘﺒﺎﻳﻦ‪.‬‬

‫ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﺘﻜﺎﻣﻠﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ )‪(p,d,q‬‬ ‫‪Autoregressive-Integrated-Moving Average Models‬‬ ‫)‪ARIMA(p,d,q‬‬ ‫ﻳﻤﻜﻦ ﻧﻤﺬﺟﺔ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﺴﺘﻘﺮة ‪ wt = ∇ d zt‬ﻋﻠﻰ ﺷﻜﻞ ﻧﻤﻮذج أﻧﺤﺪار ذاﺗﻲ‪-‬ﻣﺘﻮﺳﻂ ﻣﺘﺤﺮك ﻣﻦ‬ ‫اﻟﺪرﺟﺔ ) ‪ ( p, q‬آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫) ‪WN ( 0, σ 2‬‬

‫أو‬

‫)‬

‫‪2‬‬

‫‪φ p ( B ) wt = φ p ( B ) ∇ d zt = δ + θ q ( B ) at , at‬‬

‫‪φ p ( B )(1 − B ) zt = δ + θ q ( B ) at , at WN ( 0,σ‬‬ ‫‪d‬‬

‫وهﺬا اﻟﻨﻤﻮذج ﻳﺴﻤﻰ ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﺘﻜﺎﻣﻠﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ ) ‪( p, d , q‬‬ ‫ﺣﻴﺚ ) ∞ ‪ δ ∈ ( −∞,‬ﻣﻌﻠﻢ اﻹﻧﺠﺮاف‪.‬‬ ‫أﻣﺜﻠﺔ ﻋﻠﻰ ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﺘﻜﺎﻣﻠﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ ) ‪: ( p, d , q‬‬ ‫اوﻻ‪ :‬ﻧﻤﻮذج اﻹﻧﺤﺪاراﻟﺬاﺗﻲ‪-‬اﻟﺘﻜﺎﻣﻠﻲ ﻣﻦ اﻟﺪرﺟﺔ )‪ (١،١‬أو )‪ARIMA(١،١،٠‬‬ ‫)‪: =ARI(١،١‬‬ ‫وﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫) ‪φ1 ( B )(1 − B ) zt = δ + θ 0 ( B ) at , at WN ( 0, σ 2‬‬

‫‪(1 − φ1B )(1 − B ) zt = δ + at‬‬ ‫‪{1 − (φ1 + 1) B + φ1B 2 } zt = δ + at‬‬ ‫أي‬ ‫‪φ1 < 1‬‬

‫‪),‬‬

‫‪2‬‬

‫‪WN ( 0, σ‬‬

‫‪zt = δ + (φ1 + 1) zt −1 − φ1 zt −2 + at , at‬‬

‫ﺛﺎﻧﻴﺎ‪ :‬ﻧﻤﻮذج اﻟﺘﻜﺎﻣﻠﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ )‪(١،١‬‬ ‫أو )‪: ARIMA(٠،١،١) = IMA(١،١‬‬ ‫وﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫) ‪φ0 ( B )(1 − B ) zt = δ + θ1 ( B ) at , at WN ( 0, σ 2‬‬

‫‪WN ( 0,σ 2 ) , θ1 < 1‬‬

‫‪at‬‬

‫‪(1 − B ) zt = δ + (1 − θ1B ) at ,‬‬

‫‪WN ( 0, σ 2 ) , θ1 < 1‬‬

‫‪zt − zt −1 = δ + at − θ1at , at‬‬

‫أي‬ ‫‪٤٨‬‬

‫‪WN ( 0, σ 2 ) ,‬‬

‫‪θ1 < 1‬‬

‫‪zt = δ + zt −1 + at − θ1at , at‬‬

‫ﺛﺎﻟﺜﺎ‪ :‬ﻧﻤﻮذج اﻟﻤﺸﻲ اﻟﻌﺸﻮاﺋﻲ ﺑﺈﻧﺠﺮاف ‪ Random Walk with Trend Model‬أو‬ ‫)‪: ARIMA(٠،١،٠‬‬ ‫وﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬ ‫) ‪WN ( 0, σ 2‬‬

‫‪φ0 ( B )(1 − B ) zt = δ + θ 0 ( B ) at , at‬‬

‫) ‪WN ( 0, σ 2‬‬

‫أي‬

‫)‬

‫‪2‬‬

‫‪WN ( 0, σ‬‬

‫‪at‬‬

‫‪(1 − B ) zt = δ + at ,‬‬

‫‪zt = δ + zt −1 + at , at‬‬

‫داﻟﺔ اﻷوزان ) ‪ ψ ( B‬وﺗﻤﺜﻴﻞ ﻧﻤﺎذج )‪:ARMA(p,q‬‬ ‫ﺳﺒﻖ أن آﺘﺒﻨﺎ ﻧﻤﺎذج ) ‪ ARMA ( p, q‬ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫) ‪WN ( 0, σ 2‬‬

‫أو ﺑﺸﻜﻞ اﻹﻧﺤﺮاف ﻋﻦ اﻟﻤﺘﻮﺳﻂ‬

‫)‬

‫‪2‬‬

‫‪WN ( 0, σ‬‬

‫‪φ p ( B ) zt = δ + θ q ( B ) a t , a t‬‬ ‫‪φ p ( B )( zt − µ ) = θ q ( B ) at , at‬‬

‫ﻓﻲ آﻠﺘﺎ اﻟﺤﺎﻟﺘﻴﻦ ﺑﺎﻟﻘﺴﻤﺔ ﻋﻠﻰ ﻋﺎﻣﻞ اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ) ‪ φ p ( B‬ﻧﺠﺪ‬ ‫) ‪WN ( 0, σ 2‬‬

‫) ‪θq ( B‬‬ ‫‪a , at‬‬ ‫‪φp (B) t‬‬

‫) ‪WN ( 0, σ 2‬‬

‫‪+‬‬

‫‪δ‬‬

‫)‪φ p (1‬‬

‫) ‪θq ( B‬‬ ‫‪a , at‬‬ ‫‪φp (B) t‬‬

‫= ‪zt‬‬

‫= ‪zt − µ‬‬

‫) ‪θq ( B‬‬ ‫ﻻﺣﻆ ان ﻟﻠﻨﻤﺎذج اﻟﻤﺴﺘﻘﺮة اﻟﻨﺴﺒﺔ‬ ‫)‪φp (B‬‬ ‫‪δ‬‬ ‫ﺗﻘﻊ ﺧﺎرج داﺋﺮة اﻟﻮﺣﺪة اﻳﻀﺎ ‪= µ‬‬ ‫وﻟﻬﺬا ﺳﻮف ﻧﻜﺘﻔﻲ ﺑﺸﻜﻞ اﻹﻧﺤﺮاف ﻋﻦ اﻟﻤﺘﻮﺳﻂ ﻓﻲ‬ ‫)‪φ p (1‬‬

‫ﺗﺸﻜﻞ ﻣﺴﻠﺴﻠﺔ ﻣﺘﻘﺎرﺑﺔ وذﻟﻚ ﻷن ﺟﺬور ‪φ p ( B ) = 0‬‬

‫ﻣﻨﺎﻗﺸﺘﻨﺎ اﻟﺘﺎﻟﻴﺔ‬ ‫) ‪WN ( 0, σ 2‬‬

‫اﻟﻤﺴﻠﺴﻠﺔ اﻟﻤﺘﻘﺎرﺑﺔ‬

‫واﻟﺘﻲ ﺗﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫)‪θ (B‬‬ ‫‪zt − µ = q‬‬ ‫‪a , at‬‬ ‫‪φp (B) t‬‬ ‫) ‪θq ( B‬‬ ‫)‪φp (B‬‬

‫)‪θq ( B‬‬ ‫‪= ψ 0 B 0 + ψ 1 B1 + ψ 2 B 2 + ψ 3 B 3 + L , ψ 0 = 1‬‬ ‫)‪φp (B‬‬

‫ﺗﺴﻤﻰ داﻟﺔ اﻷوزان‪.‬‬

‫‪٤٩‬‬

‫= )‪ψ (B‬‬

‫= )‪ψ (B‬‬

‫ﺗﻌﺮﻳﻒ ‪ :١٧‬داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج ) ‪ ARMA ( p, q‬اﻟﻤﺴﺘﻘﺮ ﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬ ‫)‪θq ( B‬‬ ‫‪= ψ 0 B 0 + ψ 1 B1 + ψ 2 B 2 + ψ 3 B 3 + L , ψ 0 = 1‬‬ ‫)‪φp (B‬‬

‫= )‪ψ (B‬‬

‫∞ )‪θq ( B‬‬ ‫‪= ∑ψ B j , ψ 0 = 1‬‬ ‫‪φ p ( B ) j =0 j‬‬

‫= )‪ψ (B‬‬

‫ﺣﻴﺚ اﻷوزان هﻲ ‪ψ 0 = 1,ψ 1 ,ψ 2 ,ψ 3 ,L‬‬ ‫) ‪θq ( B‬‬ ‫‪a , at‬‬ ‫‪φp (B) t‬‬

‫ﻣﻼﺣﻈﺔ‪ :‬اﻟﻨﻤﻮذج اﻟﺬي ﻋﻠﻰ اﻟﺸﻜﻞ ) ‪WN ( 0, σ 2‬‬

‫= ‪ zt − µ‬ﻳﺴﻤﻰ ﺗﻤﺜﻴﻞ‬

‫اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك اﻟﻼﻧﻬﺎﺋﻲ ﻟﻨﻤﺎذج ) ‪. ARMA ( p, q‬‬

‫أﻣﺜﻠﺔ ﻟﺪاﻟﺔ اﻷوزان ﻟﺒﻌﺾ اﻟﻨﻤﺎذج‪:‬‬ ‫داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج )‪: AR(١‬‬ ‫ﻧﻤﻮذج )‪ AR(١‬ﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫) ‪φ1 ( B )( zt − µ ) = θ 0 ( B ) at , at WN ( 0, σ 2‬‬

‫‪(1 − φ1B )( zt − µ ) = at‬‬ ‫‪1‬‬ ‫‪a‬‬ ‫‪(1 − φ1B ) t‬‬

‫= ‪zt − µ‬‬

‫‪zt − µ = ψ ( B ) at‬‬

‫ﺣﻴﺚ‬ ‫‪1‬‬ ‫) ‪(1 − φ1B‬‬

‫= )‪ψ (B‬‬

‫ﺳﻮف ﻧﻮﺟﺪ اﻷوزان ‪ ψ 1 ,ψ 2 ,ψ 3 ,L‬ﺑﺎﻟﻄﺮﻳﻘﺔ اﻟﺘﺎﻟﻴﺔ‬ ‫‪1‬‬ ‫) ‪(1 − φ1B‬‬

‫= )‪ψ (B‬‬

‫‪ψ ( B )(1 − φ1B ) ≡ 1‬‬

‫‪B 2 + ψ 3 B 3 + L) (1 − φ1B ) ≡ 1‬‬

‫‪2‬‬

‫‪(1 + ψ B + ψ‬‬ ‫‪1‬‬

‫ﻻﺣﻆ أن اﻟﻌﻼﻗﺔ اﻷﺧﻴﺮة هﻲ ﻋﻼﻗﺔ ﺗﻜﺎﻓﺆ أي ان ﻣﻌﺎﻣﻼت ‪ B j‬ﻋﻠﻰ ﻃﺮﻓﻲ اﻟﻌﻼﻗﺔ ﻣﺘﺴﺎوﻳﺔ‪.‬‬ ‫وﺑﻤﺴﺎواة ﻣﻌﺎﻣﻼت ‪ B j‬ﻋﻠﻰ ﻃﺮﻓﻲ اﻟﻌﻼﻗﺔ ﻧﺠﺪ‬ ‫‪φ1 < 1‬‬

‫‪+ ψ 3 B 3 + L) (1 − φ1B ) ≡ 1,‬‬

‫‪2‬‬

‫‪(1 + ψ B + ψ B‬‬ ‫‪2‬‬

‫‪1‬‬

‫‪B 0 : (1)(1) ≡ 1‬‬ ‫‪B1 : ψ 1 − φ1 ≡ 0 ⇒ ψ 1 = φ1‬‬ ‫‪B 2 : ψ 2 − ψ 1φ1 ≡ 0 ⇒ ψ 2 = ψ 1φ1 = φ12‬‬ ‫‪B 3 : ψ 3 − ψ 2φ1 ≡ 0 ⇒ ψ 3 = ψ 2φ1 = φ13‬‬ ‫‪M‬‬ ‫‪B j : ψ j − ψ j −1φ1 ≡ 0 ⇒ ψ j = ψ j −1φ1 = φ1j‬‬

‫أي ان اﻷوزان ﻟﻨﻤﻮذج )‪ AR(١‬هﻲ‬ ‫‪٥٠‬‬

ψ j = φ1j , φ1 < 1

: MA(١) ‫داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج‬ ‫ ﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬MA(١) ‫ﻧﻤﻮذج‬

φ0 ( B )( zt − µ ) = θ1 ( B ) at , at WN ( 0, σ 2 )

( zt − µ ) = (1 − θ1B ) at zt − µ = ψ ( B ) at ‫ﺣﻴﺚ‬

ψ ( B ) = (1 − θ1B )

‫ ﻋﻠﻰ ﻃﺮﻓﻲ اﻟﻌﻼﻗﺔ ﻧﺠﺪ‬B j ‫ﺑﻤﺴﺎواة ﻣﻌﺎﻣﻼت‬

ψ 1 = −θ1 , ψ 2 = ψ 3 = L = 0

‫أي‬ j=0 j =1

⎧ 1, ⎪ ψ j = ⎨ −θ1 , ⎪ 0, ⎩

j≥2

φ2 ( B )( zt − µ ) = θ 0 ( B ) at , at WN ( 0,σ 2 )

(1 − φ B − φ B ) ( z − µ ) = a

: AR(٢) ‫داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج‬ ‫ ﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬AR(٢) ‫ﻧﻤﻮذج‬

2

1

zt − µ =

2

t

t

1 a (1 − φ1B − φ2 B 2 ) t

zt − µ = ψ ( B ) at

‫ﺣﻴﺚ‬ ψ (B) =

1 (1 − φ1B − φ2 B 2 )

ψ ( B ) (1 − φ1B − φ2 B ) ≡ 1

‫ ﺑﺎﻟﻄﺮﻳﻘﺔ اﻟﺴﺎﺑﻘﺔ‬ψ 1 ,ψ 2 ,ψ 3 ,L ‫و ﻧﻮﺟﺪ اﻷوزان‬

2

(1 + ψ B + ψ B 1

2

2

+ ψ 3 B 3 + L)(1 − φ1B − φ2 B 2 ) ≡ 1

B1 : ψ 1 − φ1 = 0 ⇒ ψ 1 = φ1 2 B 2 : ψ 2 − φψ 1 1 − φ2 = 0 ⇒ ψ 2 = φψ 1 1 + φ2 = φ1 + φ 2

B 3 : ψ 3 − φψ 1 2 − φ2ψ 1 = 0 ⇒ ψ 3 = φψ 1 2 + φ2ψ 1 M B j : ψ j − φψ 1 j −1 − φ2ψ j − 2 = 0 ⇒ ψ j = φψ 1 j −1 + φ 2ψ j − 2

‫ هﻲ‬AR(٢) ‫أي ان اﻷوزان ﻟﻨﻤﻮذج‬

٥١

j=0

⎧ 1, ⎪φ , ⎪ 1 ψj =⎨ 2 ⎪ φ1 + φ2 , ⎪⎩φψ 1 j −1 + φ2ψ j − 2 ,

j =1 j=2 j≥3

: MA(٢) ‫داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج‬ ‫ ﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬MA(٢) ‫ﻧﻤﻮذج‬

φ0 ( B )( zt − µ ) = θ 2 ( B ) at , at WN ( 0,σ 2 )

( zt − µ ) = (1 − θ1B − θ 2 B 2 ) at zt − µ = ψ ( B ) at ψ ( B ) = (1 − θ1B − θ 2 B

2

‫ﺣﻴﺚ‬

)

‫ ﻋﻠﻰ ﻃﺮﻓﻲ اﻟﻌﻼﻗﺔ ﻧﺠﺪ‬B j ‫ﺑﻤﺴﺎواة ﻣﻌﺎﻣﻼت‬

ψ 1 = −θ1 , ψ 2 = −θ 2 , ψ 3 = ψ 4 = ψ 5 L = 0

‫أي‬ j=0 j =1 j=2 j≥2

⎧ 1, ⎪ −θ , ⎪ ψj =⎨ 1 ⎪ −θ 2 , ⎪⎩ 0,

φ1 ( B )( zt − µ ) = θ1 ( B ) at , at

: ARMA(١،١) ‫داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج‬ ‫ ﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬ARMA(١،١) ‫ﻧﻤﻮذج‬

WN ( 0, σ 2 )

(1 − φ1B )( zt − µ ) = (1 − θ1B ) at (1 − θ1B ) a zt − µ = (1 − φ1B ) t zt − µ = ψ ( B ) a t ψ (B) =

(1 − θ1B ) (1 − φ1B )

ψ ( B )(1 − φ1B ) ≡ (1 − θ1B )

(1 + ψ B + ψ 1

‫ﺣﻴﺚ‬

2

‫ ﺑﺎﻟﻄﺮﻳﻘﺔ اﻟﺴﺎﺑﻘﺔ‬ψ 1 ,ψ 2 ,ψ 3 ,L ‫و ﻧﻮﺟﺪ اﻷوزان‬

B 2 + ψ 3 B 3 + L) (1 − φ1B ) ≡ (1 − θ1B )

B1 : ψ 1 − φ1 = −θ1 ⇒ ψ 1 = φ1 − θ1 B 2 : ψ 2 − φψ 1 1 = 0 ⇒ ψ 2 = φψ 1 1 = φ1 (φ1 − θ1 ) 2 B 3 : ψ 3 − φψ 1 2 = 0 ⇒ ψ 3 = φψ 1 2 = φ1 (φ1 − θ1 )

M

j −1 B j : ψ j − φψ (φ1 − θ1 ) 1 j −1 = 0 ⇒ ψ j = φψ 1 j −1 = φ1

‫ هﻲ‬ARMA(١،١) ‫أي ان اﻷوزان ﻟﻨﻤﻮذج‬ ٥٢

j −1 ψ j = φψ (φ1 − θ1 ) , 1 j −1 = φ1

j ≥ 1,

φ1 < 1, φ1 ≠ θ1

: ARI(١) ‫داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج‬ ‫ ﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬ARI(١) ‫ﻧﻤﻮذج‬

φ1 ( B )(1 − B )( zt − µ ) = at , at WN ( 0,σ 2 ) zt − µ =

1 a (1 − φ1B )(1 − B ) t

zt − µ = ψ ( B ) at

‫ﺣﻴﺚ‬ ψ ( B) =

1 (1 − φ1B )(1 − B )

‫ ﺑﺎﻟﻄﺮﻳﻘﺔ اﻟﺴﺎﺑﻘﺔ‬ψ 1 ,ψ 2 ,ψ 3 ,L ‫و ﻧﻮﺟﺪ اﻷوزان‬

ψ ( B )(1 − φ1B )(1 − B ) ≡ 1

(1 +ψ B +ψ (1 +ψ B +ψ 1

2

1

2

B 2 + ψ 3 B 3 + L) (1 − φ1B )(1 − B ) ≡ 1

B 2 + ψ 3 B 3 + L) (1 − (φ1 + 1) B + φ1B 2 ) ≡ 1

B1 : ψ 1 − (φ1 + 1) = 0 ⇒ ψ 1 = φ1 + 1 B 2 : ψ 2 − (φ1 + 1)ψ 1 + φ1 = 0 ⇒ ψ 2 = (φ1 + 1)ψ 1 + φ1 = (φ1 + 1) + φ1 2

B 3 : ψ 3 − (φ1 + 1)ψ 2 + φψ 1 1 = 0 ⇒ ψ 3 = (φ1 + 1)ψ 2 − φψ 1 1 M

B j : ψ j − (φ1 + 1)ψ j −1 + φψ 1 j −2 1 j − 2 = 0 ⇒ ψ j = (φ1 + 1)ψ j −1 − φψ

‫ هﻲ‬ARI(١) ‫أي ان اﻷوزان ﻟﻨﻤﻮذج‬ j=0 j =1

⎧ 1, ⎪ φ + 1, ⎪ 1 ψj =⎨ 2 ⎪ (φ1 + 1) + φ1 , ⎪(φ1 + 1)ψ j −1 − φψ 1 j −2 , ⎩

j=2 j≥3

‫وأﺧﻴﺮا ﻧﻮﺟﺪ‬ ARIMA(١،٠،١) ‫ أو‬Random Walk Mdel ‫داﻟﺔ اﻷوزان ﻟﻨﻤﻮذج اﻟﻤﺸﻲ اﻟﻌﺸﻮاﺋﻲ‬ ‫وﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬ zt = zt −1 + at , at

WN ( 0, σ 2 )

zt − zt −1 = at , at

WN ( 0, σ

2

‫أي‬

)

(1 − B ) zt = at zt =

1 a (1 − B ) t

‫ﺣﻴﺚ‬ 1 ψ ( B) = (1 − B )

٥٣

‫و ﻧﻮﺟﺪ اﻷوزان ‪ ψ 1 ,ψ 2 ,ψ 3 ,L‬ﺑﺎﻟﻄﺮﻳﻘﺔ اﻟﺴﺎﺑﻘﺔ‬

‫‪ψ ( B )(1 − B ) ≡ 1‬‬ ‫‪+ ψ 3 B 3 + L) (1 − B ) ≡ 1‬‬

‫‪2‬‬

‫‪(1 + ψ B + ψ B‬‬ ‫‪1‬‬

‫‪2‬‬

‫‪B1 : ψ 1 − 1 = 0 ⇒ ψ 1 = 1‬‬ ‫‪B 2 :ψ 2 −ψ 1 = 0 ⇒ ψ 2 = ψ 1 = 1‬‬ ‫‪B 3 :ψ 3 −ψ 2 = 0 ⇒ ψ 3 = ψ 2 = 1‬‬ ‫‪M‬‬ ‫‪B j : ψ j − ψ j −1 = 0 ⇒ ψ j = ψ j −1 = 1‬‬

‫أي ان اﻷوزان ﻟﻨﻤﻮذج اﻟﻤﺸﻲ اﻟﻌﺸﻮاﺋﻲ ) ‪ ARIMA ( 0,1,0‬هﻲ‬ ‫‪ψ j = 1,‬‬

‫‪j ≥1‬‬

‫ﺑﻌﺾ ﺧﻮاص داﻟﺔ اﻷوزان ) ‪: ψ ( B‬‬ ‫ﺳﺒﻖ أن آﺘﺒﻨﺎ ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ )‪ ARMA(p,q‬ﻋﻠﻰ اﻟﺸﻜﻞ‬ ‫) ‪WN ( 0, σ 2‬‬

‫وﺑﻜﺘﺎﺑﺔ هﺬﻩ اﻟﻌﻼﻗﻪ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫‪zt − µ = ψ ( B ) a t , a t‬‬

‫‪zt − µ = at + ψ 1at −1 + ψ 2 at −2 + ψ 3at −3 + L‬‬ ‫∞‬

‫‪= ∑ψ j at − j , ψ 0 = 1‬‬ ‫‪j =0‬‬

‫وإذا اﻓﺘﺮﺿﻨﺎ ان اﻷوزان ﺗﺘﻘﺎرب اي ∞ < ‪ψ 2j‬‬

‫∞‬

‫‪j =0‬‬

‫∑‬

‫ﻓﺈﻧﻪ ﻳﻤﻜﻦ إﺛﺒﺎت اﻟﻨﻈﺮﻳﺔ اﻟﺘﺎﻟﻴﺔ‪:‬‬

‫ﻧﻈﺮﻳﺔ ‪:١‬‬ ‫ﻟﻨﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ‪-‬اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ )‪ ARMA(p,q‬اﻟﻤﺴﺘﻘﺮ واﻟﺬي ﻳﻜﺘﺐ‬ ‫ﻋﻠﻰ اﻟﺸﻜﻞ‬ ‫∞ < ‪WN ( 0, σ 2 ) , ψ 0 = 1, ∑ j =0ψ 2j‬‬ ‫∞‬

‫∞‬

‫‪zt − µ =∑ψ j at − j , at‬‬ ‫‪j =0‬‬

‫‪ -١‬اﻟﻤﺘﻮﺳﻂ هﻮ‬

‫‪E ( zt ) = µ , ∀t‬‬

‫‪ -٢‬داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬ ‫∞‬

‫‪j +k‬‬

‫‪, k = 0,1, 2,L‬‬

‫‪∑ψ ψ‬‬ ‫‪j‬‬

‫‪j =0‬‬

‫‪2‬‬ ‫‪j‬‬

‫∞‬

‫‪∑ψ‬‬

‫= ‪ρk‬‬

‫‪j=0‬‬

‫ﻣﺜﺎل‪ :‬ﻟﻨﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ ﻣﻦ اﻟﺪرﺟﺔ )‪ AR(١‬وﺟﺪﻧﺎ ﺳﺎﺑﻘﺎ داﻟﺔ اﻷوزان‬ ‫‪ψ j = φ , φ1 < 1‬‬ ‫‪j‬‬ ‫‪1‬‬

‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ‬

‫‪٥٤‬‬

‫‪= φ1k , k = 0,1, 2,L‬‬

‫‪φ1k‬‬ ‫‪1 − φ12‬‬ ‫‪1‬‬ ‫‪1 − φ12‬‬

‫∞‬

‫=‬

‫‪∑φ1jφ1j+k‬‬ ‫‪j =0‬‬ ‫∞‬

‫‪∑φ‬‬

‫‪2j‬‬ ‫‪1‬‬

‫‪j =0‬‬

‫∞‬

‫=‬

‫‪∑ψ jψ j+k‬‬ ‫‪j =0‬‬

‫‪2‬‬ ‫‪j‬‬

‫‪∑ψ‬‬

‫وهﻲ ﻧﻔﺲ اﻟﻨﺘﻴﺠﺔ اﻟﺴﺎﺑﻘﺔ‬ ‫ﺗﻤﺮﻳﻦ‪ :‬ﺑﺄﺳﺘﺨﺪام ﻧﻈﺮﻳﺔ ‪ (٢) ٢‬أوﺟﺪ دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﻨﻤﺎذج اﻟﺘﺎﻟﻴﺔ‬ ‫)‪AR(٢), MA(١), MA(٢), ARMA(١،١), ARMA(٢،١), ARMA(١،٢‬‬

‫‪٥٥‬‬

‫∞‬

‫‪j=0‬‬

‫= ‪ρk‬‬

‫اﻟﻔﺼﻞ اﻟﺮاﺑﻊ‬ ‫اﻟﺘﻨﺒﺆات ذات ﻣﺘﻮﺳﻂ ﻣﺮﺑﻊ اﻟﺨﻄﺄ اﻷدﻧﻰ ﻟﻨﻤﺎذج )‪ARMA(p,q‬‬ ‫‪Minimum Mean Square Error Forecasts for‬‬ ‫‪ARMA(p,q) Models‬‬ ‫ﻓﻲ ﺍﻟﻔﻘﺭﺓ ﺍﻟﺴﺎﺒﻘﺔ ﻜﺘﺒﻨﺎ ﻨﻤﻭﺫﺝ ﺍﻹﻨﺤﺩﺍﺭ ﺍﻟﺫﺍﺘﻲ‪-‬ﺍﻟﻤﺘﻭﺴﻁ ﺍﻟﻤﺘﺤﺭﻙ ﻤﻥ ﺍﻟﺩﺭﺠﺔ )‪ARMA(p,q‬‬ ‫ﺍﻟﻤﺴﺘﻘﺭ ﻋﻠﻰ ﺍﻟﺸﻜل‬ ‫∞ < ‪WN ( 0, σ 2 ) , ψ 0 = 1, ∑ j =0ψ 2j‬‬ ‫∞‬

‫∞‬

‫‪zt − µ =∑ψ j at − j , at‬‬ ‫‪j =0‬‬

‫ﺃﻭ‬

‫‪zt − µ = at + ψ 1at −1 + ψ 2at −2 + ψ 3at −3 +L‬‬ ‫∞‬

‫‪=∑ψ j at − j , ψ 0 = 1‬‬ ‫‪j =0‬‬

‫ﻤﻼﺤﻅﺔ‪ :‬ﻫﺫﺍ ﻴﻨﻁﺒﻕ ﺃﻴﻀﺎ ﻋﻠﻰ ﻨﻤﺎﺫﺝ )‪ ARIMA(p,d,q‬ﺒﺸﻜل ﻋﺎﻡ‪.‬‬ ‫ﻟﻤﺘﺴﻠﺴﻠﺔ ﺯﻤﻨﻴﺔ ﻤﺸﺎﻫﺩﺓ } ‪ {z1 , z2 ,L, zn −1 , zn‬ﺍﻟﺘﻨﺒﺅﺍﺕ ‪ zn ( l ) , l ≥ 1‬ﻟﻠﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ‬ ‫‪ zn +l , l ≥ 1‬ﻴﻤﻜﻥ ﺍﻥ ﺘﻜﺘﺏ ﻋﻠﻰ ﺍﻟﺸﻜل‬ ‫‪zn ( l ) = ξ 0an + ξ1an −1 + ξ 2an −2 + L, l ≥ 1‬‬

‫ﺍﻟﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn +l , l ≥ 1‬ﺘﻜﺘﺏ ﺒﺩﻻﻟﺔ ﺍﻟﻨﻤﻭﺫﺝ ﻜﺎﻟﺘﺎﻟﻲ‬

‫‪zn + l − µ = an +l + ψ 1an + l−1 + L + ψ l−1an +1 + ψ l an + ψ l+1an −1 + L, l ≥ 1‬‬

‫ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻟﺨﻁﺄ ﻴﻌﻁﻰ ﺒﺎﻟﻌﻼﻗﺔ )ﺃﻨﻅﺭ ﺘﻌﺭﻴﻑ ‪( ٥‬‬

‫‪2‬‬

‫⎦⎤‪E ⎡⎣ zn + l − zn ( l )⎤⎦ = E ⎡⎣ an + l + ψ 1an + l−1 + L + ψ l−1an +1 + (ψ l − ξ 0 ) an + (ψ l+1 − ξ1 ) an −1 + L‬‬ ‫‪2‬‬

‫∞‬

‫‪= (1 + ψ 12 + L + ψ l2−1 ) σ 2 + ∑ (ψ l+ j − ξ j ) σ 2‬‬ ‫‪2‬‬

‫‪j =0‬‬

‫ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻟﺨﻁﺄ ﺍﻷﺩﻨﻰ ﻴﻨﺘﺞ ﻤﻥ ﺘﺼﻐﻴﺭ ﺍﻟﻌﻼﻗﺔ ﺍﻟﺴﺎﺒﻘﺔ ﺒﺎﻟﻨﺴﺒﺔ ﻟﻸﻭﺯﺍﻥ ‪ ξ j‬ﻟﺠﻤﻴﻊ ﻗﻴﻡ ‪j‬‬ ‫ﻭﻫﺫﺍ ﻴﻤﻜﻥ ﺇﺫﺍ ﻭﻓﻘﻁ ﺇﺫﺍ ﺤﻘﻘﺕ ﺍﻷﻭﺯﺍﻥ ‪ ξ j‬ﺍﻟﻌﻼﻗﺔ ﺍﻟﺘﺎﻟﻴﺔ‬ ‫‪j = 0,1, 2,L , l ≥ 1‬‬

‫‪ξ j = ψ l+ j ,‬‬

‫ﻭﻋﻠﻴﻪ ﻓﺈﻥ ﺍﻟﺘﻨﺒﺅﺍﺕ ﺫﺍﺕ ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻟﺨﻁﺄ ﺍﻷﺩﻨﻰ ‪ MMSE Forecasts‬ﺘﻌﻁﻰ ﺒﺎﻟﻌﻼﻗﺔ‬ ‫‪zn ( l ) = ψ l an + ψ l+1an −1 + ψ l+2an −2 + L, l ≥ 1‬‬

‫ﻨﻅﺭﻴﺔ ‪:٢‬‬ ‫ﺃﺨﻁﺎﺀ ﺍﻟﺘﻨﺒﺅ ﺘﻌﻁﻰ ﺒﺎﻟﻌﻼﻗﺔ‪:‬‬

‫‪en ( l ) = zn +l − zn ( l ) = an +l + ψ 1an +l−1 + ψ 2an +l−2 + L + ψ l−1an+1 , l ≥ 1‬‬

‫ﻭﺘﺒﺎﻴﻥ ﺃﺨﻁﺎﺀ ﺍﻟﺘﻨﺒﺅ ﺘﻌﻁﻰ ﺒﺎﻟﻌﻼﻗﺔ‪:‬‬ ‫‪V ⎡⎣ en ( l )⎤⎦ = σ 2 (1 + ψ 12 + ψ 22 + L + ψ l2−1 ) , l ≥ 1‬‬

‫‪٥٦‬‬

‫ﺍﻟﺼﻴﻐﺔ ‪ zn ( l ) = ψ lan + ψ l+1an−1 + ψ l+2an −2 + L, l ≥ 1‬ﻏﻴﺭ ﻋﻤﻠﻴﺔ ﻹﻴﺠﺎﺩ ﺍﻟﺘﻨﺒﺅﺍﺕ ﻟﻠﻘﻴﻡ‬ ‫ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn+l , l ≥ 1‬ﻭﺫﻟﻙ ﻷﻨﻨﺎ ﻨﺤﺘﺎﺝ ﺇﻟﻰ ﻤﻌﺭﻓﺔ ﺍﻟﻘﻴﻡ } ‪. {a1 , a2 ,L, an−1 , an‬‬

‫ﺘﻌﺭﻴﻑ ‪ : ١٨‬ﻤﺠﻤﻭﻋﺔ ﺍﻟﻤﻌﻠﻭﻤﺎﺕ )} ‪ I ({z1 , z2 ,L , zn −1 , zn‬ﺘﻜﺎﻓﺊ ﻤﺠﻤﻭﻋﺔ ﺍﻟﻤﻌﻠﻭﻤﺎﺕ‬ ‫)} ‪ I ({a1 , a2 ,L , an −1 , an‬ﻭﺫﻟﻙ ﺒﺎﻟﻤﻌﻨﻰ ﺃﻥ ﺍﻟﻤﺠﻤﻭﻋﺔ } ‪ {a1 , a2 ,L, an−1 , an‬ﺘﺤﺘﻭﻯ ﻋﻠﻰ ﻨﻔﺱ‬ ‫ﺍﻟﻤﻌﻠﻭﻤﺎﺕ ﻋﻥ ﺍﻟﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ } ‪. {z1 , z2 ,L, zn −1 , zn‬‬ ‫ﻤﻼﺤﻅﺔ‪ :‬ﺍﻟﻤﺘﺴﻠﺴﺔ ﺍﻟﺯﻤﻨﻴﺔ } ‪ {z1 , z2 ,L, zn −1 , zn‬ﻴﻤﻜﻥ ﻤﺸﺎﻫﺩﺘﻬﺎ ﻭﻗﻴﺎﺴﻬﺎ ﻭﻟﻜﻥ ﺍﻟﻤﺘﻠﺴﻠﺔ‬ ‫} ‪ {a1 , a2 ,L, an−1 , an‬ﻻﻴﻤﻜﻥ ﻤﺸﺎﻫﺩﺘﻬﺎ ﺃﻭ ﻗﻴﺎﺴﻬﺎ‪.‬‬

‫ﻨﻅﺭﻴﺔ ‪ : ٣‬ﺍﻟﻤﺘﻨﺒﺊ ﺫﺍ ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻟﺨﻁﺄ ﺍﻷﺩﻨﻰ ‪ MMSE Forecasts‬ﻴﻌﻁﻰ ﺒﺎﻟﻌﻼﻗﺔ‬ ‫‪zn ( l ) = E ( zn + l zn , zn −1 ,L) , l ≥ 1‬‬ ‫ﺃﻱ ﻫﻭ ﺍﻟﺘﻭﻗﻊ ﺍﻟﺸﺭﻁﻲ ﻟﻠﻘﻴﻤﺔ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn +l , l ≥ 1‬ﻤﻌﻁﻰ } ‪. {z1 , z2 ,L, zn −1 , zn‬‬ ‫ﺘﺴﺘﺨﺩﻡ ﻨﻅﺭﻴﺔ ‪ ٢‬ﻋﻤﻠﻴﺎ ﻹﻴﺠﺎﺩ ﻗﻴﻡ ﺍﻟﺘﻨﺒﺅﺍﺕ ﺒﺩﻻ ﻤﻥ ﺍﻟﺼﻴﻐﺔ‬ ‫‪zn ( l ) = ψ l an + ψ l+1an −1 + ψ l+2an −2 + L, l ≥ 1‬‬

‫ﻭﺫﻟﻙ ﺘﺒﻌﺎ ﻟﻠﻤﻼﺤﻅﺔ ﺍﻟﺴﺎﺒﻘﺔ‪.‬‬

‫ﻗﺎﻋﺩﺓ ‪:٢‬‬ ‫‪⎧a , j ≤ 0‬‬ ‫‪1 − E ( an + j zn , zn −1 ,L) = ⎨ n + j‬‬ ‫‪j>0‬‬ ‫‪⎩ 0,‬‬ ‫‪j≤0‬‬ ‫‪⎧ zn + j ,‬‬ ‫⎨ = )‪2 − E ( zn + j zn , zn −1 ,L‬‬ ‫‪⎩ zn ( j ) , j > 0‬‬

‫ﻨﻅﺭﻴﺔ ‪ ٣‬ﻤﻊ ﺍﻟﻘﺎﻋﺩﺓ ‪ ٢‬ﺘﻌﻁﻲ ﻁﺭﻴﻘﺔ ﻋﻤﻠﻴﺔ ﻭﺴﻬﻠﺔ ﻹﻴﺠﺎﺩ ﺘﻨﺒﺅﺍﺕ ﻟﻠﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ‬ ‫‪zn + l , l ≥ 1‬‬

‫ﺍﻟﺩﺍﻟﺔ ‪ zn ( l ) , l ≥ 1‬ﻜﺩﺍﻟﺔ ﻟﺯﻤﻥ ﺍﻟﺘﻘﺩﻡ‬

‫‪l ≥1‬‬

‫ﺘﻌﺭﻴﻑ ‪:١٩‬‬ ‫ﻋﻨﺩ ﻨﻘﻁﺔ ﺍﻻﺼل ﻟﻠﺯﻤﻥ ‪ n‬ﺘﺴﻤﻰ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ‪.‬‬

‫دوال اﻟﺘﻨﺒﺆ ﻟﻨﻤﺎذج )‪: ARIMA(p,d,q‬‬ ‫ﺍﻭﻻ‪ :‬ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﻟﻨﻤﻭﺫﺝ )‪: AR(١‬‬

‫ﻟﻨﻔﺘﺭﺽ ﺍﻨﻨﺎ ﺸﺎﻫﺩﻨﺎ ﺍﻟﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ } ‪ {z1 , z2 ,L, zn −1 , zn‬ﺤﺘﻰ ﺍﻟﺯﻤﻥ ‪ n‬ﻭﺍﻟﺘﻲ ﻨﻌﺘﻘﺩ ﺍﻨﻬﺎ‬ ‫ﺘﺘﺒﻊ ﻨﻤﻭﺫﺝ )‪ AR(١‬ﻭﺍﻟﺫﻱ ﻴﻜﺘﺏ ﻋﻠﻰ ﺍﻟﺸﻜل‬

‫‪٥٧‬‬

‫) ∞ ‪WN ( 0, σ 2 ) , φ1 < 1, µ ∈ ( −∞,‬‬

‫‪zt − µ = φ1 ( zt −1 − µ ) + at , at‬‬

‫ﻨﺭﻴﺩ ﺃﻥ ﻨﺘﻨﺒﺄ ﻋﻥ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn +1 , zn +2 , zn+3 ,L‬ﺃﻭ ﺒﺸﻜل ﻋﺎﻡ ‪. zn +l , l ≥ 1‬‬ ‫ﻤﻥ ﻨﻅﺭﻴﺔ ‪ ٣‬ﻨﺠﺩ‬ ‫‪zn ( l ) = E ( zn + l zn , zn −1 ,L) , l ≥ 1‬‬

‫‪=µ +E ⎡⎣ ⎡⎣φ1 ( zn + l−1 − µ ) + an +l ⎤⎦ zn , zn −1 ,L⎤⎦ , l ≥ 1‬‬

‫‪=µ +E ⎡⎣φ1 ( zn + l−1 − µ ) zn , zn −1 ,L + an + l zn , zn −1 ,L⎤⎦ , l ≥ 1‬‬

‫‪= µ +φ1E ⎡⎣( zn + l−1 zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an + l zn , zn −1 ,L⎤⎦ , l ≥ 1‬‬

‫ﺃﻱ‬ ‫‪zn ( l ) = µ +φ1E ⎣⎡ ( zn + l−1 zn , zn −1 ,L) − µ ⎦⎤ + E ⎡⎣ an + l zn , zn −1 ,L⎤⎦ , l ≥ 1‬‬

‫ﻨﺤل ﻫﺫﻩ ﺍﻟﻌﻼﻗﺔ ﺘﻜﺭﺍﺭﻴﺎ ﻭﺒﺈﺴﺘﺨﺩﺍﻡ ﺍﻟﻘﺎﻋﺩﺓ ‪٢‬‬

‫⎦⎤‪l = 1: zn (1) = µ +φ1E ⎣⎡ ( zn zn , zn −1 ,L) − µ ⎦⎤ + E ⎡⎣ an +1 zn , zn −1 ,L‬‬ ‫) ‪= µ +φ1 ( zn − µ‬‬

‫⎦⎤‪l = 2 : zn ( 2 ) = µ +φ1E ⎡⎣( zn +1 zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an + 2 zn , zn −1 ,L‬‬ ‫⎦⎤ ‪= µ +φ1 ⎡⎣ zn (1) − µ‬‬

‫⎦⎤‪l = 3 : zn ( 3) = µ +φ1E ⎡⎣( zn + 2 zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an +3 zn , zn −1 ,L‬‬ ‫⎦⎤ ‪= µ +φ1 ⎡⎣ zn ( 2 ) − µ‬‬

‫ﻭﻫﻜﺫﺍ ﺒﺸﻜل ﻋﺎﻡ‬

‫‪zn ( l ) = µ +φ1 ⎡⎣ zn ( l − 1) − µ ⎤⎦ , l ≥ 1‬‬

‫ﻭﻫﻲ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﺫﺍﺕ ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻷﺨﻁﺎﺀ ﺍﻷﺩﻨﻰ ﻟﻨﻤﻭﺫﺝ )‪AR(١‬‬

‫ﺘﻌﺭﻴﻑ ‪:٢٠‬‬ ‫ﺸﺭﻁ ﺍﻹﺴﺘﻤﺭﺍﺭ ‪ Continuity Condition‬ﻴﺘﻁﻠﺏ ﺃﻨﻪ ﻋﻨﺩﻤﺎ ﺘﻜﻭﻥ ‪ l = 1‬ﻓﺈﻥ‬ ‫‪zn ( l − 1) = zn ( 0 ) = zn‬‬

‫ﻤﻥ ﻨﻅﺭﻴﺔ ‪ ٢‬ﺘﺒﺎﻴﻥ ﺃﺨﻁﺎﺀ ﺍﻟﺘﻨﺒﺅ ﺘﻌﻁﻰ ﻤﻥ ﺍﻟﻌﻼﻗﺔ‬ ‫‪V ⎡⎣ en ( l )⎤⎦ = σ 2 (1 + ψ 12 + ψ 22 + L + ψ l2−1 ) , l ≥ 1‬‬

‫ﺴﺒﻕ ﺃﻥ ﺍﺸﺘﻘﻘﻨﺎ ﺩﺍﻟﺔ ﺍﻷﻭﺯﺍﻥ ﻟﻨﻤﻭﺫﺝ )‪ AR(١‬ﻭﻫﻲ‬ ‫‪ψ j = φ1j , φ1 < 1‬‬

‫‪), l ≥1‬‬

‫ﻭﺒﺎﻟﺘﻌﻭﻴﺽ ﻓﻲ ﺼﻴﻐﺔ ﺍﻟﺘﺒﺎﻴﻥ ﻨﺠﺩ‬

‫(‬

‫( ‪V ⎡⎣ en ( l )⎤⎦ = σ 2 1 + φ12 + φ14 + L + φ1‬‬

‫)‪2 l −1‬‬

‫‪1 − φ12 l‬‬ ‫‪, l ≥1‬‬ ‫‪1 − φ12‬‬

‫‪=σ2‬‬

‫ﻤﺜﺎل‪ :‬ﻤﺘﺴﻠﺴﻠﺔ ﺯﻤﻨﻴﺔ ﻤﺸﺎﻫﺩﻩ ﻭﺠﺩ ﺍﻨﻬﺎ ﺘﺘﺒﻊ ﺍﻟﻨﻤﻭﺫﺝ‬ ‫‪٥٨‬‬

‫) ‪WN ( 0,0.024‬‬

‫‪zt − 0.97 = 0.85 ( zt −1 − 0.97 ) + at , at‬‬

‫ﺇﺫﺍ ﻜﺎﻨﺕ ﺍﻟﻤﺸﺎﻫﺩﺓ ﺍﻷﺨﻴﺭﺓ ﻫﻲ ‪ ، z156 = 0.49‬ﺃﻭﺠﺩ ﺘﻨﺒﺅﺍﺕ ﻟﻠﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪z157 , z158 , z159‬‬

‫ﻭﺃﻭﺠﺩ ﺃﺨﻁﺎﺀ ﺍﻟﺘﻨﺒﺅ ﻟﻬﺎ‪.‬‬ ‫ﺍﻟﺤل‪ :‬ﻤﻥ ﺍﻟﺼﻴﻐﺔ ‪ zn ( l ) = µ +φ1 ⎡⎣ zn ( l − 1) − µ ⎤⎦ , l ≥ 1‬ﻨﺠﺩ‬ ‫) ‪z156 (1) = 0.97+0.85 ( z156 − 0.97‬‬

‫‪= 0.97+0.85 ( 0.49 − 0.97 ) = 0.56‬‬ ‫) ‪z156 ( 2 ) = 0.97+0.85 ( z156 (1) − 0.97‬‬ ‫‪= 0.97+0.85 ( 0.56 − 0.97 ) = 0.62‬‬ ‫) ‪z156 ( 3) = 0.97+0.85 ( z156 ( 2 ) − 0.97‬‬ ‫‪= 0.97+0.85 ( 0.62 − 0.97 ) = 0.68‬‬

‫ﻭﺍﻟﺘﺒﺎﻴﻨﺎﺕ‬ ‫‪1 − φ12 l‬‬ ‫‪, l ≥1‬‬ ‫‪1 − φ12‬‬

‫‪V ⎡⎣ en ( l )⎤⎦ = σ 2‬‬

‫‪V ⎡⎣ e156 (1)⎤⎦ = 0.024‬‬ ‫‪= 0.041‬‬ ‫‪= 0.054‬‬

‫‪4‬‬

‫)‪1 − ( 0.85‬‬

‫‪2‬‬

‫)‪1 − ( 0.85‬‬

‫‪6‬‬

‫)‪1 − ( 0.85‬‬

‫‪2‬‬

‫)‪1 − ( 0.85‬‬

‫‪V ⎡⎣ e156 ( 2 )⎤⎦ = 0.024‬‬ ‫‪V ⎡⎣ e156 ( 2 )⎤⎦ = 0.024‬‬

‫ﺜﺎﻨﻴﺎ‪ :‬ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﻟﻨﻤﻭﺫﺝ )‪: AR(٢‬‬

‫ﻟﻨﻔﺘﺭﺽ ﺍﻨﻨﺎ ﺸﺎﻫﺩﻨﺎ ﺍﻟﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ } ‪ {z1 , z2 ,L, zn −1 , zn‬ﺤﺘﻰ ﺍﻟﺯﻤﻥ ‪ n‬ﻭﺍﻟﺘﻲ ﻨﻌﺘﻘﺩ ﺍﻨﻬﺎ‬ ‫ﺘﺘﺒﻊ ﻨﻤﻭﺫﺝ )‪ AR(٢‬ﻭﺍﻟﺫﻱ ﻴﻜﺘﺏ ﻋﻠﻰ ﺍﻟﺸﻜل‬ ‫‪WN ( 0, σ 2 ) , µ ∈ ( −∞, ∞ ) ,‬‬

‫‪zt = µ + φ1 ( zt −1 − µ ) + φ2 ( zt −2 − µ ) + at , at‬‬

‫‪φ2 − φ1 < 1, φ2 + φ1 < 1, φ2 < 1‬‬

‫ﻨﺭﻴﺩ ﺃﻥ ﻨﺘﻨﺒﺄ ﻋﻥ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn +1 , zn +2 , zn+3 ,L‬ﺃﻭ ﺒﺸﻜل ﻋﺎﻡ ‪. zn +l , l ≥ 1‬‬ ‫ﻤﻥ ﻨﻅﺭﻴﺔ ‪ ٣‬ﻨﺠﺩ‬ ‫‪zn ( l ) = E ( zn + l zn , zn −1 ,L) , l ≥ 1‬‬

‫‪=µ +E ⎡⎣ ⎡⎣φ1 ( zn + l−1 − µ ) + φ2 ( zn + l−2 − µ ) + an + l ⎤⎦ zn , zn −1 ,L⎤⎦ , l ≥ 1‬‬

‫‪=µ +E ⎡⎣φ1 ( zn + l−1 − µ ) zn , zn −1 ,L + φ2 ( zn + l−2 − µ ) zn , zn −1 ,L + an + l zn , zn −1 ,L⎤⎦ , l ≥ 1‬‬

‫‪= µ +φ1E ⎡⎣( zn +l−1 zn , zn −1 ,L) − µ ⎤⎦ + φ2 E ⎡⎣( zn + l−2 zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an + l zn , zn −1 ,L⎤⎦ , l ≥ 1‬‬

‫ﺃﻱ‬

‫‪٥٩‬‬

zn ( l ) = µ +φ1E ⎡⎣ ( zn + l −1 zn , zn −1 ,L) − µ ⎤⎦ + φ2 E ⎡⎣ ( zn + l −2 zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an + l zn , zn −1 ,L⎤⎦ , l ≥ 1

٢ ‫ﻨﺤل ﻫﺫﻩ ﺍﻟﻌﻼﻗﺔ ﺘﻜﺭﺍﺭﻴﺎ ﻭﺒﺈﺴﺘﺨﺩﺍﻡ ﺍﻟﻘﺎﻋﺩﺓ‬

l = 1: zn (1) = µ +φ1E ⎡⎣( zn zn , zn −1 ,L) − µ ⎤⎦ + φ2 E ⎡⎣( zn −1 zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an +1 zn , zn −1 ,L⎤⎦ = µ +φ1 ( zn − µ ) + φ2 ( zn −1 − µ )

l = 2 : zn ( 2 ) = µ +φ1E ⎡⎣( zn +1 zn , zn −1 ,L) − µ ⎤⎦ + φ2 E ⎡⎣ ( zn zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an +2 zn , zn −1 ,L⎤⎦ = µ +φ1 ⎡⎣ zn (1) − µ ⎤⎦ + φ2 ( zn − µ )

l = 3 : zn ( 3) = µ +φ1E ⎡⎣( zn + 2 zn , zn −1 ,L) − µ ⎤⎦ + φ2 E ⎡⎣( zn +1 zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an +3 zn , zn −1 ,L⎤⎦ = µ +φ1 ⎡⎣ zn ( 2 ) − µ ⎤⎦ +φ2 ⎡⎣ zn (1) − µ ⎤⎦

l = 4 : zn ( 4 ) = µ +φ1E ⎡⎣( zn +3 zn , zn −1 ,L) − µ ⎤⎦ + φ2 E ⎡⎣( zn + 2 zn , zn −1 ,L) − µ ⎤⎦ + E ⎡⎣ an + 4 zn , zn −1 ,L⎤⎦ = µ +φ1 ⎡⎣ zn ( 3) − µ ⎤⎦ +φ2 ⎡⎣ zn ( 2 ) − µ ⎤⎦

‫ﻭﻫﻜﺫﺍ ﺒﺸﻜل ﻋﺎﻡ‬

zn ( l ) = µ +φ1 ⎡⎣ zn ( l − 1) − µ ⎤⎦ + φ2 ⎡⎣ zn ( l − 2 ) − µ ⎤⎦ , l ≥ 1

AR(٢) ‫ﻭﻫﻲ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﺫﺍﺕ ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻷﺨﻁﺎﺀ ﺍﻷﺩﻨﻰ ﻟﻨﻤﻭﺫﺝ‬ ‫ ﻭﺩﺍﻟﺔ ﺍﻷﻭﺯﺍﻥ‬٢ ‫ﻭﻴﻤﻜﻥ ﺤﺴﺎﺏ ﺘﺒﺎﻴﻨﺎﺕ ﺃﺨﻁﺎﺀ ﺍﻟﺘﻨﺒﺅ ﻤﻥ ﻨﻅﺭﻴﺔ‬ ⎧ 1, ⎪φ , ⎪ 1 ψj =⎨ 2 ⎪ φ1 + φ2 , ⎪⎩φψ 1 j −1 + φ2ψ j − 2 ,

j=0 j =1 j=2 j≥3

: ARIMA(٠،١،١) ‫ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﻟﻨﻤﻭﺫﺝ‬:‫ﺜﺎﻟﺜﺎ‬

‫ ﻭﺍﻟﺘﻲ ﻨﻌﺘﻘﺩ ﺍﻨﻬﺎ‬n ‫{ ﺤﺘﻰ ﺍﻟﺯﻤﻥ‬z1 , z2 ,L, zn −1 , zn } ‫ﻟﻨﻔﺘﺭﺽ ﺍﻨﻨﺎ ﺸﺎﻫﺩﻨﺎ ﺍﻟﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ‬ ‫ ﻭﺍﻟﺫﻱ ﻴﻜﺘﺏ ﻋﻠﻰ ﺍﻟﺸﻜل‬ARIMA(٠،١،١) ‫ﺘﺘﺒﻊ ﻨﻤﻭﺫﺝ‬ zt = zt −1 + at − θ1at −1 , at

WN ( 0, σ 2 )

. zn+l , l ≥ 1 ‫ ﺃﻭ ﺒﺸﻜل ﻋﺎﻡ‬zn +1 , zn +2 , zn+3 ,L ‫ﻨﺭﻴﺩ ﺃﻥ ﻨﺘﻨﺒﺄ ﻋﻥ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ‬ ‫ ﻨﺠﺩ‬٣ ‫ﻤﻥ ﻨﻅﺭﻴﺔ‬ zn ( l ) = E ( zn + l zn , zn −1 ,L) , l ≥ 1

=E ( zn +l−1 zn , zn −1 ,L) + E ( an +l zn , zn −1 ,L) − θ1E ( an +l−1 zn , zn −1 ,L) , l ≥ 1

٢ ‫ﻨﺤل ﻫﺫﻩ ﺍﻟﻌﻼﻗﺔ ﺘﻜﺭﺍﺭﻴﺎ ﻭﺒﺈﺴﺘﺨﺩﺍﻡ ﺍﻟﻘﺎﻋﺩﺓ‬

٦٠

‫‪zn ( l ) =E ( zn + l−1 zn , zn −1 ,L) + E ( an + l zn , zn −1 ,L) − θ1E ( an + l−1 zn , zn −1 ,L) , l ≥ 1‬‬

‫)‪l = 1: zn (1) =E ( zn zn , zn −1 ,L) + E ( an +1 zn , zn −1 ,L) − θ1E ( an zn , zn −1 ,L‬‬ ‫‪= zn − θ1an‬‬

‫)‪l = 2 : zn ( 2 ) =E ( zn +1 zn , zn −1 ,L) + E ( an + 2 zn , zn −1 ,L) − θ1E ( an +1 zn , zn −1 ,L‬‬ ‫)‪= zn (1‬‬ ‫)‪l = 3 : zn ( 3) =E ( zn + 2 zn , zn −1 ,L) + E ( an +3 zn , zn −1 ,L) − θ1E ( an +1 zn , zn −1 ,L‬‬ ‫) ‪= zn ( 2‬‬

‫ﻭﻫﻜﺫﺍ ﺒﺸﻜل ﻋﺎﻡ‬

‫‪zn ( l ) = zn ( l − 1) , l ≥ 2‬‬

‫ﻭﻫﻜﺫﺍ ﻓﺈﻥ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﺫﺍﺕ ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻷﺨﻁﺎﺀ ﺍﻷﺩﻨﻰ ﻟﻨﻤﻭﺫﺝ )‪ ARIMA(٠،١،١‬ﺘﻌﻁﻰ‬ ‫ﺒﺎﻟﻌﻼﻗﺔ‬ ‫‪⎧ zn − θ1an , l = 1‬‬ ‫⎨ = ) ‪zn ( l‬‬ ‫‪⎩ zn ( l − 1) , l > 1‬‬

‫ﺭﺍﺒﻌﺎ ‪ :‬ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﻟﻨﻤﻭﺫﺝ )‪: MA(١‬‬

‫ﻟﻨﻔﺘﺭﺽ ﺍﻨﻨﺎ ﺸﺎﻫﺩﻨﺎ ﺍﻟﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ } ‪ {z1 , z2 ,L, zn −1 , zn‬ﺤﺘﻰ ﺍﻟﺯﻤﻥ ‪ n‬ﻭﺍﻟﺘﻲ ﻨﻌﺘﻘﺩ ﺍﻨﻬﺎ‬ ‫ﺘﺘﺒﻊ ﻨﻤﻭﺫﺝ )‪ MA(١‬ﻭﺍﻟﺫﻱ ﻴﻜﺘﺏ ﻋﻠﻰ ﺍﻟﺸﻜل‬ ‫) ‪WN ( 0, σ 2‬‬

‫‪zt = µ + at − θ1at −1 , at‬‬

‫ﻨﺭﻴﺩ ﺃﻥ ﻨﺘﻨﺒﺄ ﻋﻥ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn +1 , zn +2 , zn+3 ,L‬ﺃﻭ ﺒﺸﻜل ﻋﺎﻡ ‪. zn +l , l ≥ 1‬‬ ‫ﻤﻥ ﻨﻅﺭﻴﺔ ‪ ٣‬ﻨﺠﺩ‬ ‫‪zn ( l ) = E ( zn + l zn , zn −1 ,L) , l ≥ 1‬‬

‫‪=µ + E ( an + l zn , zn −1 ,L) − θ1E ( an +l−1 zn , zn −1 ,L) , l ≥ 1‬‬

‫ﻨﺤل ﻫﺫﻩ ﺍﻟﻌﻼﻗﺔ ﺘﻜﺭﺍﺭﻴﺎ ﻭﺒﺈﺴﺘﺨﺩﺍﻡ ﺍﻟﻘﺎﻋﺩﺓ ‪٢‬‬

‫‪zn ( l ) =µ + E ( an + l zn , zn −1 ,L) − θ1E ( an + l−1 zn , zn −1 ,L) , l ≥ 1‬‬

‫)‪l = 1: zn (1) =µ + E ( an +1 zn , zn −1 ,L) − θ1E ( an zn , zn −1 ,L‬‬ ‫‪= µ − θ1an‬‬

‫)‪l = 2 : zn ( 2 ) =µ + E ( an + 2 zn , zn −1 ,L) − θ1E ( an +1 zn , zn −1 ,L‬‬ ‫‪=µ‬‬ ‫)‪l = 3 : zn ( 3) =µ + E ( an +3 zn , zn −1 ,L) − θ1E ( an + 2 zn , zn −1 ,L‬‬ ‫‪=µ‬‬

‫ﻭﻫﻜﺫﺍ ﺒﺸﻜل ﻋﺎﻡ‬

‫‪zn ( l ) = µ , l ≥ 2‬‬

‫ﻭﻫﻜﺫﺍ ﻓﺈﻥ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﺫﺍﺕ ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻷﺨﻁﺎﺀ ﺍﻷﺩﻨﻰ ﻟﻨﻤﻭﺫﺝ )‪ MA(١‬ﺘﻌﻁﻰ ﺒﺎﻟﻌﻼﻗﺔ‬ ‫‪⎧ µ − θ1an , l = 1‬‬ ‫⎨ = ) ‪zn ( l‬‬ ‫‪l≥2‬‬ ‫‪⎩ µ,‬‬

‫ﺨﺎﻤﺴﺎ ‪ :‬ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﻟﻨﻤﻭﺫﺝ )‪: MA(٢‬‬

‫‪٦١‬‬

‫ﻟﻨﻔﺘﺭﺽ ﺍﻨﻨﺎ ﺸﺎﻫﺩﻨﺎ ﺍﻟﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ } ‪ {z1 , z2 ,L, zn −1 , zn‬ﺤﺘﻰ ﺍﻟﺯﻤﻥ ‪ n‬ﻭﺍﻟﺘﻲ ﻨﻌﺘﻘﺩ ﺍﻨﻬﺎ‬ ‫ﺘﺘﺒﻊ ﻨﻤﻭﺫﺝ )‪ MA(٢‬ﻭﺍﻟﺫﻱ ﻴﻜﺘﺏ ﻋﻠﻰ ﺍﻟﺸﻜل‬ ‫) ‪WN ( 0, σ 2‬‬

‫‪zt = µ + at − θ1at −1 − θ 2 at − 2 , at‬‬

‫ﻨﺭﻴﺩ ﺃﻥ ﻨﺘﻨﺒﺄ ﻋﻥ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn +1 , zn +2 , zn+3 ,L‬ﺃﻭ ﺒﺸﻜل ﻋﺎﻡ ‪. zn+l , l ≥ 1‬‬ ‫ﻤﻥ ﻨﻅﺭﻴﺔ ‪ ٣‬ﻨﺠﺩ‬ ‫‪zn ( l ) = E ( zn + l zn , zn −1 ,L) , l ≥ 1‬‬

‫‪=µ + E ( an + l zn , zn −1 ,L) − θ1E ( an +l−1 zn , zn −1 ,L) − θ 2 E ( an +l−2 zn , zn −1 ,L) , l ≥ 1‬‬

‫ﻨﺤل ﻫﺫﻩ ﺍﻟﻌﻼﻗﺔ ﺘﻜﺭﺍﺭﻴﺎ ﻭﺒﺈﺴﺘﺨﺩﺍﻡ ﺍﻟﻘﺎﻋﺩﺓ ‪٢‬‬

‫‪zn ( l ) =µ + E ( an + l zn , zn −1 ,L) − θ1E ( an + l−1 zn , zn −1 ,L) − θ 2 E ( an + l−2 zn , zn −1 ,L) , l ≥ 1‬‬

‫)‪l = 1: zn (1) =µ + E ( an +1 zn , zn −1 ,L) − θ1E ( an zn , zn −1 ,L) − θ 2 E ( an −1 zn , zn −1 ,L‬‬ ‫‪= µ − θ1an − θ 2an −1‬‬

‫)‪l = 2 : zn ( 2 ) =µ + E ( an + 2 zn , zn −1 ,L) − θ1E ( an +1 zn , zn −1 ,L) − θ 2 E ( an zn , zn −1 ,L‬‬ ‫‪= µ − θ 2 an‬‬

‫)‪l = 3 : zn ( 3) =µ + E ( an +3 zn , zn −1 ,L) − θ1E ( an + 2 zn , zn −1 ,L) − θ 2 E ( an +1 zn , zn −1 ,L‬‬ ‫‪=µ‬‬

‫ﻭﻫﻜﺫﺍ ﺒﺸﻜل ﻋﺎﻡ‬

‫‪zn ( l ) = µ , l ≥ 3‬‬

‫ﻭﻫﻜﺫﺍ ﻓﺈﻥ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﺫﺍﺕ ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻷﺨﻁﺎﺀ ﺍﻷﺩﻨﻰ ﻟﻨﻤﻭﺫﺝ )‪ MA(٢‬ﺘﻌﻁﻰ ﺒﺎﻟﻌﻼﻗﺔ‬ ‫‪⎧ µ − θ1an − θ 2an −1 , l = 1‬‬ ‫⎪‬ ‫‪zn ( l ) = ⎨ µ − θ 2 an ,‬‬ ‫‪l=2‬‬ ‫‪⎪ µ,‬‬ ‫‪l≥3‬‬ ‫⎩‬

‫ﺴﺎﺩﺴﺎ ‪ :‬ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﻟﻨﻤﻭﺫﺝ )‪: ARMA(١،١‬‬

‫ﻟﻨﻔﺘﺭﺽ ﺍﻨﻨﺎ ﺸﺎﻫﺩﻨﺎ ﺍﻟﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ } ‪ {z1 , z2 ,L, zn −1 , zn‬ﺤﺘﻰ ﺍﻟﺯﻤﻥ ‪ n‬ﻭﺍﻟﺘﻲ ﻨﻌﺘﻘﺩ ﺍﻨﻬﺎ‬ ‫ﺘﺘﺒﻊ ﻨﻤﻭﺫﺝ )‪ ARMA(١،١‬ﻭﺍﻟﺫﻱ ﻴﻜﺘﺏ ﻋﻠﻰ ﺍﻟﺸﻜل‬ ‫‪WN ( 0, σ 2 ) , φ1 ≠ θ1 , φ1 < 1‬‬

‫‪zt = µ + φ1 ( zt −1 − µ ) + at − θ1at −1 , at‬‬

‫ﻨﺭﻴﺩ ﺃﻥ ﻨﺘﻨﺒﺄ ﻋﻥ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn +1 , zn +2 , zn+3 ,L‬ﺃﻭ ﺒﺸﻜل ﻋﺎﻡ ‪. zn +l , l ≥ 1‬‬ ‫ﻤﻥ ﻨﻅﺭﻴﺔ ‪ ٣‬ﻨﺠﺩ‬ ‫‪zn ( l ) = E ( zn + l zn , zn −1 ,L) , l ≥ 1‬‬

‫‪=µ + φ1E ⎡⎣( zn + l−1 − µ ) zn , zn −1 ,L⎤⎦ + E ( an + l zn , zn −1 ,L) − θ1E ( an + l−1 zn , zn −1 ,L) , l ≥ 1‬‬

‫ﻨﺤل ﻫﺫﻩ ﺍﻟﻌﻼﻗﺔ ﺘﻜﺭﺍﺭﻴﺎ ﻭﺒﺈﺴﺘﺨﺩﺍﻡ ﺍﻟﻘﺎﻋﺩﺓ ‪٢‬‬

‫‪٦٢‬‬

‫‪zn ( l ) =µ + φ1 E ⎡⎣( zn + l−1 − µ ) zn , zn −1 ,L⎤⎦ + E ( an +l zn , zn −1 ,L) − θ1E ( an + l−1 zn , zn −1 ,L) , l ≥ 1‬‬

‫)‪l = 1: zn (1) =µ + φ1E ⎣⎡( zn − µ ) zn , zn −1 ,L⎦⎤ + E ( an +1 zn , zn −1 ,L) − θ1E ( an zn , zn −1 ,L‬‬ ‫‪= µ + φ1 ( zn − µ ) − θ1an‬‬

‫)‪l = 2 : zn ( 2 ) =µ + φ1 E ⎡⎣( zn +1 − µ ) zn , zn −1 ,L⎤⎦ + E ( an +2 zn , zn −1 ,L) − θ1E ( an +1 zn , zn −1 ,L‬‬ ‫⎦⎤ ‪= µ + φ1 ⎡⎣ zn (1) − µ‬‬

‫)‪l = 3 : zn ( 3) =µ + φ1 E ⎡⎣( zn + 2 − µ ) zn , zn −1 ,L⎤⎦ + E ( an +3 zn , zn −1 ,L) − θ1E ( an + 2 zn , zn −1 ,L‬‬ ‫⎦⎤ ‪= µ + φ1 ⎡⎣ zn ( 2 ) − µ‬‬

‫ﻭﻫﻜﺫﺍ ﺒﺸﻜل ﻋﺎﻡ‬

‫‪zn ( l ) = µ + φ1 ⎡⎣ zn ( l − 1) − µ ⎤⎦ , l ≥ 2‬‬

‫ﻭﻫﻜﺫﺍ ﻓﺈﻥ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﺫﺍﺕ ﻤﺘﻭﺴﻁ ﻤﺭﺒﻊ ﺍﻷﺨﻁﺎﺀ ﺍﻷﺩﻨﻰ ﻟﻨﻤﻭﺫﺝ )‪ ARMA(١،١‬ﺘﻌﻁﻰ‬ ‫ﺒﺎﻟﻌﻼﻗﺔ‬ ‫‪l =1‬‬ ‫‪l≥2‬‬

‫‪⎧⎪ µ + φ1 ( zn − µ ) − θ1an ,‬‬ ‫⎨ = ) ‪zn ( l‬‬ ‫‪⎪⎩ µ + φ1 ⎡⎣ zn ( l − 1) − µ ⎤⎦ ,‬‬

‫ﺘﻤﺭﻴﻥ‪:‬‬ ‫ﻟﻨﻤﻭﺫﺝ )‪ ARMA(١،١‬ﻭﺍﻟﺫﻱ ﻴﻜﺘﺏ ﻋﻠﻰ ﺍﻟﺸﻜل‬

‫‪WN ( 0, σ 2 ) , φ1 ≠ θ1 , φ1 < 1‬‬

‫‪zt = µ + φ1 ( zt −1 − µ ) + at − θ1at −1 , at‬‬

‫ﺒﺭﻫﻥ ﺍﻥ ﻋﻨﺩﻤﺎ ﺘﺅﻭل ‪ φ1 → 1‬ﻓﺈﻥ )‪ ARMA(1,1) → IMA(1,1‬ﻭﻤﻥ ﺜﻡ ﺃﻭﺠﺩ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ‬ ‫ﻟﻨﻤﻭﺫﺝ )‪. IMA(١،١‬‬ ‫ﺘﻤﺭﻴﻥ‪:‬‬ ‫ﺃﻭﺠﺩ ﺩﻭﺍل ﺍﻟﺘﻨﺒﺅ ﻭﺘﺒﺎﻴﻥ ﺃﺨﻁﺎﺀ ﺍﻟﺘﻨﺒﺅ ﻟﻜل ﻤﻥ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﺘﺎﻟﻴﺔ‪:‬‬ ‫‪ARIMA(١،١،١), ARIMA(٢،١،٠), ARIMA(٠،١،٢), ARIMA(١،٢،٠),‬‬ ‫‪ARIMA(٠،٢،١), ARIMA(٠،٢،٠).‬‬ ‫ﺤﺩﻭﺩ ﺍﻟﺘﻨﺒﺅ ‪: Forecasting Limits‬‬ ‫ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ‪ zn ( l ) , l ≥ 1‬ﻋﻨﺩ ﻗﻴﻤﺔ ﻤﻌﻴﻨﺔ ﺘﻌﻁﻲ ﻤﺎﻴﺴﻤﻰ ﺒﺘﻨﺒﺅ ﺍﻟﻨﻘﻁﺔ ‪ Point Forecast‬ﻭﺍﻟﺫﻱ‬ ‫ﻻﻴﻜﻔﻲ ﺍﻭ ﻴﻔﻴﺩ ﻓﻲ ﺇﺘﺨﺎﺫ ﻗﺭﺍﺭﺍﺕ ﺇﺤﺼﺎﺌﻴﺔ ﻋﻥ ﺍﻟﻅﺎﻫﺭﺓ ﺍﻟﻌﺸﻭﺍﺌﻴﺔ ﺍﻟﻤﺩﺭﻭﺴﺔ ﻷﻥ‬ ‫‪P ( Z n + m = zn ( m ) ) = 0, for some m > 0‬‬

‫ﺃﻱ ﺃﻥ ﻤﻘﺩﺍﺭ ﺘﺄﻜﺩﻨﺎ ) ﺃﻭ ﺇﺤﺘﻤﺎل( ﻤﻥ ﺃﻥ ﺍﻟﻘﻴﻤﺔ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﺍﻟﻤﺭﺍﺩ ﺍﻟﺘﻨﺒﺅ ﻋﻨﻬﺎ ﺘﺴﺎﻭﻱ ﺍﻟﻘﻴﻤﺔ‬ ‫ﺍﻟﻤﻌﻁﺎﺓ ﻤﻥ ﺩﺍﻟﺔ ﺍﻟﺘﻨﺒﺅ ﺘﺴﺎﻭﻱ ﺍﻟﺼﻔﺭ ﺃﻱ ﺍﻨﻨﺎ ﻏﻴﺭ ﻤﺘﺄﻜﺩﻴﻥ ﺇﻁﻼﻗﺎ ﻭﺒﺎﻟﺘﺎﻟﻲ ﻻﻓﺎﺌﺩﺓ ﻤﻥ ﺍﻟﺘﻨﺒﺅ‪.‬‬ ‫ﻟﻠﺘﻐﻠﺏ ﻋﻠﻰ ﺫﻟﻙ ﻭﺃﻹﺴﺘﻔﺎﺩﺓ ﻤﻥ ﺍﻟﺘﻨﺒﺅﺍﺕ ﻨﺴﺘﺨﺩﻡ ﻤﺎﻴﺴﻤﻰ ﺒﺘﻨﺒﺅ ﺍﻟﻔﺘﺭﺓ ‪Interval Forecast‬‬ ‫ﻭﻫﻲ ﻋﺒﺎﺭﺓ ﻋﻥ ﻓﺘﺭﺓ ﻤﺜل ]‪ [a, b‬ﻋﻠﻰ ﺨﻁ ﺍﻷﻋﺩﺍﺩ ﺍﻟﺤﻘﻴﻘﻴﺔ ﺒﺤﻴﺙ ﻴﻜﻭﻥ‬ ‫) ‪P ( a ≤ Z n +m ≤ b ) = (1 − α‬‬

‫ﻭﺒﻬﺫﺍ ﻨﺴﺘﻁﻴﻊ ﺃﻥ ﻨﺤﺩﺩ ﺩﺭﺠﺔ ﺘﺄﻜﺩﻨﺎ ﻤﻥ ﺃﻥ ﺍﻟﻘﻴﻤﺔ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﺍﻟﻤﺭﺍﺩ ﺍﻟﺘﻨﺒﺅ ﻋﻨﻬﺎ ﺘﻘﻊ ﺒﻴﻥ ﺍﻟﻘﻴﻡ ‪a‬‬ ‫ﻭ ‪ b‬ﺒﺩﺭﺠﺔ ﺘﺄﻜﺩ ﺃﻭ ﺇﺤﺘﻤﺎل ‪ ) 1 − α‬ﺃﻭ ‪ ( 100 × (1 − α ) %‬ﻓﻤﺜﻼ ﻟﻭ ﻜﺎﻨﺕ ‪ α = 0.05‬ﻓﺈﻨﻨﺎ‬ ‫ﻨﻜﻭﻥ ﻤﺘﺄﻜﺩﻴﻥ ﻭﺒﺈﺤﺘﻤﺎل ‪ ٩٥٪‬ﺍﻥ ﺍﻟﻘﻴﻤﺔ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﺘﻘﻊ ﺒﻴﻥ ﺍﻟﻘﻴﻡ ‪ a‬ﻭ ‪. b‬‬

‫‪٦٣‬‬

‫ﺘﻌﺭﻴﻑ ‪ :٢١‬ﻋﻠﻰ ﺇﻓﺘﺭﺍﺽ ﺃﻥ ) ‪ at N ( 0,σ 2‬ﻓﺈﻥ ﺤﺩﻭﺩ ‪ 100 × (1 − α ) %‬ﻓﺘﺭﺓ ﺘﻨﺒﺅ ﻟﻠﻘﻴﻤﺔ‬ ‫ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn+l , l ≥ 1‬ﺘﻌﻁﻰ ﺒﺎﻟﻌﻼﻗﺔ‬

‫{‬

‫}‬

‫‪12‬‬

‫⎦⎤) ‪zn ( l ) ± uα 2 V ⎡⎣ en ( l‬‬ ‫‪α‬‬ ‫⎟⎞ ‪ 100 ⎛⎜ 1 −‬ﻟﻠﺘﻭﺯﻴﻊ )‪. N ( 0,1‬‬ ‫⎠‪2‬‬ ‫⎝‬

‫ﺤﻴﺙ ‪ uα 2‬ﺍﻟﻤﺌﻴﻥ‬

‫ﻓﻤﺜﻼ ﻋﻨﺩﻤﺎ ‪ α = 0.05‬ﻓﺈﻥ ‪. u0.025 = 1.96‬‬ ‫ﻤﻼﺤﻅﺔ‪ :‬ﻓﻲ ﺍﻟﺘﻌﺭﻴﻑ ﺇﻓﺘﺭﻀﻨﺎ ﺃﻥ ﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﻀﺠﺔ ﺍﻟﺒﻴﻀﺎﺀ ) ‪ at N ( 0, σ 2‬ﻭﻫﺫﺍ ﻤﻤﻜﻥ‬ ‫ﺇﻋﺘﻤﺎﺩﺍ ﻋﻠﻰ ﻨﻅﺭﻴﺔ ﻨﻬﺎﻴﺔ ﻤﺭﻜﺯﻴﺔ‪.‬‬ ‫ﻤﺜﺎل‪ :‬ﻤﺘﺴﻠﺴﻠﺔ ﺯﻤﻨﻴﺔ ﻤﺸﺎﻫﺩﻩ ﻭﺠﺩ ﺍﻨﻬﺎ ﺘﺘﺒﻊ ﺍﻟﻨﻤﻭﺫﺝ‬ ‫) ‪N ( 0,0.024‬‬

‫‪zt − 0.97 = 0.85 ( zt −1 − 0.97 ) + at , at‬‬

‫ﺇﺫﺍ ﻜﺎﻨﺕ ﺍﻟﻤﺸﺎﻫﺩﺓ ﺍﻷﺨﻴﺭﺓ ﻫﻲ ‪ ، z156 = 0.49‬ﺃﻭﺠﺩ ﺘﻨﺒﺅﺍﺕ ﻟﻠﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪z157 , z158 , z159‬‬

‫ﻭﺃﻭﺠﺩ ﺃﺨﻁﺎﺀ ﺍﻟﺘﻨﺒﺅ ﻟﻬﺎ ﻭﻤﻥ ﺜﻡ ﺃﻭﺠﺩ ﻓﺘﺭﺍﺕ ﺘﻨﺒﺅ ‪ ٩٥٪‬ﻟﻠﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ‪.‬‬ ‫ﺍﻟﺤل‪ :‬ﺴﺒﻕ ﺃﻥ ﺤﺴﺒﻨﺎ ﻓﻲ ﻤﺜﺎل ﺴﺎﺒﻕ ﺍﻟﺘﻨﺒﺅﺍﺕ ﻭ ﺃﺨﻁﺎﺀ ﺍﻟﺘﻨﺒﺅ ﻜﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫ﺍﻟﺘﻨﺒﺅﺍﺕ‬ ‫‪z156 (1) = 0.56, z156 ( 2 ) = 0.62, z156 ( 3) = 0.68‬‬

‫ﻭﺍﻟﺘﺒﺎﻴﻨﺎﺕ‬ ‫‪V ⎡⎣ e156 (1)⎤⎦ = 0.024, V ⎡⎣ e156 ( 2 )⎤⎦ = 0.041, V ⎡⎣ e156 ( 2 )⎤⎦ = 0.054‬‬ ‫ﻓﺘﺭﺍﺕ ﺘﻨﺒﺅ ‪ ٩٥٪‬ﻟﻠﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ z157 , z158 , z159‬ﻨﻭﺠﺩﻫﺎ ﻤﻥ ﺼﻴﻐﺔ ﺘﻌﺭﻴﻑ ‪٢١‬‬

‫}‬

‫‪12‬‬

‫‪= 0.56 ± 1.96 0.024 = 0.56 ± 0.304‬‬

‫}‬

‫‪12‬‬

‫{‬

‫⎦⎤) ‪zn ( l ) ± uα 2 V ⎡⎣ en ( l‬‬

‫{‬ ‫‪{V ⎡⎣e‬‬ ‫‪{V ⎡⎣e‬‬

‫⎦⎤)‪1 − z156 (1) ± u0.025 V ⎡⎣ e156 (1‬‬

‫‪= 0.62 ± 1.96 0.041 = 0.62 ± 0.397‬‬

‫}⎦⎤) ‪( 2‬‬

‫‪156‬‬

‫‪2 − z156 ( 2 ) ± u0.025‬‬

‫‪= 0.68 ± 1.96 0.054 = 0.68 ± 0.455‬‬

‫}⎦⎤)‪( 3‬‬

‫‪156‬‬

‫‪3 − z156 ( 3) ± u0.025‬‬

‫‪12‬‬

‫‪12‬‬

‫ﺃﻱ ﺃﻥ ) ‪ z157 ∈ ( 0.256,0.864‬ﺒﺈﺤﺘﻤﺎل ‪ ٠,٩٥‬ﻭ ﻭﻜﺫﻟﻙ ) ‪ z158 ∈ ( 0.223,1.017‬ﻭ ﻜﺫﻟﻙ ﺃﻴﻀﺎ‬ ‫)‪. z159 ∈ ( 0.225,1.135‬‬

‫‪٦٤‬‬

٦٥

‫اﻟﻔﺼﻞ اﻟﺨﺎﻣﺲ‬ ‫ﺗﺼﻤﻴﻢ وﺑﻨﺎء ﻧﻈﺎم ﺗﻨﺒﺆ إﺣﺼﺎﺋﻲ ‪Designing and Building‬‬ ‫‪: Statistical Forecasting System‬‬ ‫ﺳﺒﻖ أن ذآﺮﻧﺎ ان اﻟﺨﻄﻮة اﻷوﻟﻰ ﻟﺘﺼﻤﻴﻢ ﻧﻈﺎم ﺗﻨﺒﺆ هﻲ ﺑﻨﺎء ﻧﻤﻮذج ‪.‬إن ﻋﻤﻠﻴﺔ ﺑﻨﺎء ﻧﻤﻮذج‬ ‫إﺣﺼﺎﺋﻲ هﻲ ﻋﻤﻠﻴﺔ ﺗﻜﺮارﻳﺔ ‪ Iterative‬ﺗﺘﻜﻮن ﻣﻦ ﺗﺤﺪﻳﺪ اﻟﻨﻤﻮذج ‪ ،‬ﺗﻘﺪﻳﺮ اﻟﻨﻤﻮذج )وﻧﻘﺼﺪ ﺑﻬﺎ‬ ‫ﺗﻘﺪﻳﺮ ﻣﻌﺎﻟﻢ اﻟﻨﻤﻮذج( و إﺧﺘﺒﺎر اﻟﻨﻤﻮذج‪.‬‬ ‫ﺗﻌﻴﻴﻦ أو ﺗﺤﺪﻳﺪ اﻟﻨﻤﻮذج )‪: Model Identification (Specification‬‬ ‫ﻓﻲ ﻣﺮﺣﻠﺔ ﺗﺤﺪﻳﺪ اﻟﻨﻤﻮذج ﻧﺴﺘﺨﺪم اﻟﺒﻴﺎﻧﺎت أو اﻟﻤﺸﺎهﺪات اﻟﺴﺎﺑﻘﺔ ) اﻟﺘﺎرﻳﺦ( واي ﻣﻌﻠﻮﻣﺎت اﺧﺮى‬ ‫ﻋﻦ اﻟﻜﻴﻔﻴﺔ اﻟﺘﻲ ﺗﻮﻟﺪت ﺑﻬﺎ اﻟﻤﺘﺴﻠﺴﻠﺔ وذﻟﻚ ﻹﻗﺘﺮاح ﻣﺠﻤﻮﻋﺔ ﻣﻦ اﻟﻨﻤﺎذج اﻟﻤﻨﺎﺳﺒﺔ‪ .‬وﻳﺘﻢ ﺗﻌﻴﻴﻦ أو‬ ‫ﺗﺤﺪﻳﺪ اﻟﻨﻤﻮذج ﺣﺴﺐ اﻟﺨﻄﻮات اﻟﻌﺮﻳﻀﺔ اﻟﺘﺎﻟﻴﺔ‪:‬‬ ‫اﻟﺨﻄﻮة اﻻوﻟﻰ‪ :‬ﺗﺤﻮﻳﻞ ﺗﺜﺒﻴﺖ اﻟﺘﺒﺎﻳﻦ ‪: Variance-stabilizing Transformation‬‬ ‫ﺑﻌﺪ رﺳﻢ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻓﻲ ﻣﺨﻄﻂ زﻣﻨﻲ ‪ Time Plot‬وإﺟﺮاء ﺑﻌﺾ اﻹﺧﺘﺒﺎرات اﻹﺣﺼﺎﺋﻴﺔ ﻟﻤﻌﺮﻓﺔ‬ ‫ﻓﻴﻤﺎ إذا آﺎن اﻟﺘﺒﺎﻳﻦ ﺛﺎﺑﺖ‪ ،‬وﻓﻲ ﺣﺎﻟﺔ ﻋﺪم ﺛﺒﺎت اﻟﺘﺒﺎﻳﻦ او إذا آﺎن اﻟﺘﺒﺎﻳﻦ ﻳﺘﻐﻴﺮ ﻣﻊ ﻣﺴﺘﻮى‬ ‫اﻟﻤﺘﺴﻠﺴﻠﺔ ﻓﺈﻧﻨﺎ ﻧﻄﺒﻖ اﻟﺘﺤﻮﻳﻞ اﻟﻠﻮﻏﺎرﺗﻤﻲ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ وﻧﻔﺤﺼﻬﺎ ﻣﻦ ﺟﺪﻳﺪ ﻓﺈذا ﺗﻢ ﺗﺜﺒﻴﺖ‬ ‫اﻟﺘﺒﺎﻳﻦ وإﻻ ﻧﻠﺠﺄ إﻟﻲ ﺗﻄﺒﻴﻖ أﺣﺪ اﻟﺘﺤﻮﻳﻼت اﻟﺘﻲ ذآﺮﻧﺎهﺎ ﻓﻲ ﺟﺪول ﺻﻔﺤﺔ ‪.٤١‬‬ ‫اﻟﺨﻄﻮة اﻟﺜﺎﻧﻴﺔ‪ :‬إﺧﺘﻴﺎر درﺟﺔ اﻟﺘﻔﺮﻳﻖ ‪: d‬‬ ‫إذا آﺎﻧﺖ اﻟﻤﺘﻠﺴﻠﺴﺔ أو ﺗﺤﻮﻳﻠﻬﺎ ﻏﻴﺮ ﻣﺴﺘﻘﺮة ﻓﻲ اﻟﻤﺘﻮﺳﻂ ﻓﻴﺠﺐ ﻋﻠﻴﻨﺎ ﺗﺤﺪﻳﺪ درﺟﺔ اﻟﺘﻔﺮﻳﻖ ‪ d‬اﻟﺘﻲ‬ ‫ﺗﺠﻌﻞ اﻟﻤﻤﺘﻠﺴﻠﺴﺔ أو ﺗﺤﻮﻳﻠﻬﺎ ﻣﺴﺘﻘﺮة ﻓﻲ اﻟﻤﺘﻮﺳﻂ وﻧﻘﻮم ﺑﺄﺧﺬ اﻟﺘﻔﺮﻳﻖ اﻷول ﺛﻢ ﻧﻔﺤﺺ اﻟﺘﺎﻟﻲ‪:‬‬ ‫‪ -١‬اﻟﻤﺨﻄﻄﺎت اﻟﺰﻣﻨﻴﺔ ﻟﻠﻤﻤﺘﻠﺴﻠﺴﺔ أو ﺗﺤﻮﻳﻠﻬﺎ‪.‬‬ ‫‪ -٢‬ﻣﺨﻄﻄﺎت داﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﻌﻴﻨﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻲ ‪ SACF‬و ‪. SPACF‬‬ ‫‪ -٣‬إﺟﺮاء ﺗﻔﺮﻳﻖ ﺁﺧﺮ إذا اﺣﺘﺎج اﻷﻣﺮ وإﻋﺎدة اﻟﺨﻄﻮات ‪ ١‬و ‪ ٢‬اﻟﺴﺎﺑﻘﺘﻴﻦ‪.‬‬ ‫اﻟﻤﺨﻄﻄﺎت اﻟﺰﻣﻨﻴﺔ ﻟﻠﻤﺘﺴﻠﺴﻼت ﻏﻴﺮ اﻟﻤﺴﺘﻘﺮة ﺗﺒﻴﻦ ﺗﻐﻴﺮ ﻓﻲ اﻟﻤﺴﺘﻮى وداﻟﺔ ﺗﺮاﺑﻂ ذاﺗﻲ ﻋﻴﻨﻲ‬ ‫ﻣﺘﺨﺎﻣﺪة ﺑﺒﻂء آﻤﺎ ان داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻲ ﺗﻌﻄﻲ ﻗﻴﻤﺔ واﺣﺪة ﻗﺮﻳﺒﺔ ﻣﻦ اﻟﻮاﺣﺪ‬ ‫اﻟﺼﺤﻴﺢ )ﺑﻐﺾ اﻟﻨﻈﺮ ﻋﻦ اﻹﺷﺎرة( وﺑﻘﻴﺔ اﻟﻘﻴﻢ ﻗﺮﻳﺒﺔ ﺟﺪا ﻣﻦ اﻟﺼﻔﺮ‪.‬‬ ‫ﻣﻼﺣﻈﺔ‪ :‬درﺟﺔ اﻟﺘﻔﺮﻳﻖ ‪ d‬ﻏﺎﻟﺒﺎ ﻣﺎ ﺗﻜﻮن ‪ ٠‬او ‪ ١‬او ‪. ٢‬‬ ‫اﻟﺨﻄﻮة اﻟﺜﺎﻟﺜﺔ‪ :‬ﺗﺤﺪﻳﺪ ‪ p‬و ‪: q‬‬ ‫ﺑﻌﺪ ان ﻧﺤﺼﻞ ﻋﻠﻰ ﻣﺘﺴﻠﺴﻠﺔ ﻣﺴﺘﻘﺮة ﻓﻲ آﻞ ﻣﻦ اﻟﺘﺒﺎﻳﻦ واﻟﻤﺘﻮﺳﻂ ﻧﻘﻮم ﺑﺘﺤﺪﻳﺪ درﺟﺔ اﻹﻧﺤﺪار‬ ‫اﻟﺬاﺗﻲ ‪ p‬ودرﺟﺔ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ‪ q‬وذﻟﻚ ﺑﻤﻘﺎرﻧﺔ أﻧﻤﺎط داﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﻌﻴﻨﻲ واﻟﺘﺮاﺑﻂ‬ ‫اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻲ ﻣﻊ اﻷﻧﻤﺎط اﻟﻨﻈﺮﻳﺔ ﻟﺪاﻟﺘﻲ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬ ‫ﻣﺴﺘﺮﺷﺪﻳﻦ ﺑﺨﻮاص ﻧﻤﺎذج )‪ ARMA(p,q‬اﻟﻤﺬآﻮرة ﻓﻲ ﺻﻔﺤﺔ ‪ ٣٨‬واﻟﺠﺪول اﻟﺘﺎﻟﻲ ﻳﻌﻄﻲ هﺬﻩ‬ ‫اﻟﺨﻮاص ﻟﺒﻌﺾ اﻟﻨﻤﺎذج اﻟﺸﺎﺋﻌﺔ‪:‬‬

‫اﻟﻨﻤﻮذج‬ ‫)‪ (١,d,٠‬و)‪AR(١‬‬ ‫)‪ (٢,d,٠‬و)‪AR(٢‬‬ ‫)‪ (p,d,٠‬و)‪AR(p‬‬ ‫)‪ (٠,d,١‬و )‪MA(١‬‬ ‫)‪ (٠,d,٢‬و )‪MA(٢‬‬

‫‪ACF‬‬ ‫ﺗﺨﺎﻣﺪ اﺳﻲ أواﺳﻲ ﻣﺘﺮدد‬ ‫ﺗﺨﺎﻣﺪ اﺳﻲ او ﺗﺨﺎﻣﺪ ﺟﻴﺒﻲ‬ ‫ﺗﺨﺎﻣﺪ اﺳﻲ و‪ /‬او ﺗﺨﺎﻣﺪ ﺟﻴﺒﻲ‬ ‫‪ρ k = 0, k > 1‬‬

‫‪PACF‬‬ ‫‪φkk = 0, k > 1‬‬ ‫‪φkk = 0, k > 2‬‬ ‫‪φkk = 0, k > p‬‬

‫ﻳﺴﻴﻄﺮ ﻋﻠﻴﻬﺎ ﺗﺨﺎﻣﺪ اﺳﻲ‬ ‫ﻳﺴﻴﻄﺮ ﻋﻠﻴﻬﺎ ﺗﺨﺎﻣﺪ اﺳﻲ او ﺟﻴﺒﻲ‬

‫‪ρ k = 0, k > 2‬‬

‫‪٦٦‬‬

‫)‪ (٠,d,q‬و )‪MA(q‬‬ ‫)‪ (١,d,١‬و )‪ARMA(١،١‬‬ ‫)‪ (p,d,q‬و )‪ARMA(p,q‬‬

‫ﻳﺴﻴﻄﺮ ﻋﻠﻴﻬﺎ ﺗﺨﺎﻣﺪ اﺳﻲ و‪ /‬او ﺟﻴﺒﻲ‬ ‫‪ρ k = 0, k > q‬‬ ‫ﺗﺘﻨﺎﻗﺺ وﻳﺴﻴﻄﺮ ﻋﻠﻴﻬﺎ ﺗﺨﺎﻣﺪ اﺳﻲ‬ ‫ﺗﺘﻨﺎﻗﺺ وﺗﺘﺨﺎﻣﺪ اﺳﻴﺎ‬ ‫ﻣﻦ اﻟﺘﺨﻠﻒ ‪١‬‬ ‫ﻣﻦ اﻟﺘﺨﻠﻒ ‪١‬‬ ‫ﺗﺘﻨﺎﻗﺺ ﺑﻌﺪ اﻟﺘﺨﻠﻒ‬ ‫ﺗﺘﻨﺎﻗﺺ ﺑﻌﺪ اﻟﺘﺨﻠﻒ ‪ q - p‬وﺗﺘﺨﺎﻣﺪ‬ ‫‪ p – q‬وﻳﺴﻴﻄﺮ‬ ‫اﺳﻴﺎ و ‪ /‬او ﺟﻴﺒﻴﺎ ﺑﻌﺪ اﻟﺘﺨﻠﻒ ‪q – p‬‬ ‫ﻋﻠﻴﻬﺎ ﺗﺨﺎﻣﺪ اﺳﻲ‬ ‫و‪/‬او ﺟﻴﺒﻲ ﺑﻌﺪ‬ ‫اﻟﺘﺨﻠﻒ ‪p – q‬‬

‫اﻟﺨﻄﻮة اﻟﺮاﺑﻌﺔ‪ :‬إﺿﺎﻓﺔ ﻣﻌﻠﻢ إﻧﺠﺮاف‪:‬‬ ‫إذا آﺎﻧﺖ اﻟﻤﺘﺴﻠﺴﺔ ﺗﺤﺘﺎج إﻟﻰ ﺗﻔﺮﻳﻖ ﻓﻴﺠﺐ ﻋﻠﻴﻨﺎ اﻟﺘﺄآﺪ ﻓﻴﻤﺎ إذا آﺎن ﻋﻠﻴﻨﺎ إﺿﺎﻓﺔ إﻧﺠﺮاف ﻣﻌﻠﻮم‬ ‫‪ δ‬إﻟﻰ اﻟﻨﻤﻮذج وهﺬا ﻳﺘﻢ ﺑﻤﻘﺎرﻧﺔ ﻣﺘﻮﺳﻂ اﻟﻌﻴﻨﺔ ‪ w‬ﻟﻠﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﻔﺮﻗﺔ اﻟﻤﺴﺘﻘﺮة ﻣﻊ اﻟﺨﻄﺄ‬ ‫اﻟﻤﻌﻴﺎري ﻟﻬﺬا اﻟﻤﺘﻮﺳﻂ‬ ‫‪12‬‬

‫‪⎡c‬‬ ‫⎤‬ ‫⎥) ‪s.e ( w ) ≅ ⎢ 0 (1 + 2r1 + 2r2 + L + 2rK‬‬ ‫‪⎣n‬‬ ‫⎦‬

‫ﺣﻴﺚ ‪ c0 = γˆ0‬و ‪ r1 ,L, rK‬هﻲ اﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﻌﻴﻨﻴﺔ اﻟﻤﻌﻨﻮﻳﺔ ﻟﻠﺪرﺟﺔ ‪ . K‬وﻳﻜﻮن اﻹﺧﺘﺒﺎر‬

‫‪H0 : δ = 0‬‬ ‫‪H1 : δ ± 0‬‬

‫‪w‬‬ ‫وﻧﺮﻓﺾ ‪ H 0‬ﻋﻨﺪ ‪ α = 0.05‬إذاآﺎﻧﺖ ‪> 1.96‬‬ ‫) ‪s.e ( w‬‬

‫‪.‬‬

‫ﺗﻘﺪﻳﺮ اﻟﻨﻤﻮذج ‪: Model Estimation‬‬ ‫ﺑﻌﺪ ﺗﺤﺪﻳﺪ ﺷﻜﻞ اﻟﻨﻤﻮذج ﻻﺑﺪ ﻣﻦ ﺗﻘﺪﻳﺮ ﻣﻌﺎﻟﻢ اﻟﻨﻤﻮذج ‪ δ‬و ‪ φ1 ,K ,φ p‬و ‪ θ1 ,K ,θ q‬و ‪ σ‬وذﻟﻚ‬ ‫ﺑﺈﺳﺘﺨﺪام اﻟﺒﻴﺎﻧﺎت اﻟﺘﺎرﻳﺨﻴﺔ اﻟﻤﺘﻮﻓﺮة ﻟﺪﻳﻨﺎ‪.‬‬ ‫ﻟﻨﻔﺘﺮض ان ﻟﺪﻳﻨﺎ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ اﻟﻤﺸﺎهﺪة ‪ z1 , z2 ,K, zn −1 , zn‬واﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح‬ ‫‪2‬‬

‫أو‬

‫) ‪N ( 0, σ 2‬‬

‫‪φ p ( B ) wt = δ + θ q ( B ) at , at‬‬

‫‪N ( 0, σ‬‬

‫‪φ p ( B ) zt = δ + θ q ( B ) at , at‬‬

‫)‬

‫‪2‬‬

‫ﺣﻴﺚ ) ‪ φ p ( B‬و ) ‪ θ q ( B‬ﻻﻳﻮﺟﺪ ﺑﻴﻨﻬﺎ ﺟﺬور ﻣﺸﺘﺮآﺔ وﺟﺬور اﻟﻤﻌﺎدﻟﺔ ‪ φ p ( B ) = 0‬ﺗﻘﻊ ﺟﻤﻴﻌﻬﺎ‬ ‫ﺧﺎرج داﺋﺮة اﻟﻮﺣﺪة ) ﺷﺮط اﻹﺳﺘﻘﺮار(‪.‬‬ ‫هﻨﺎك ﻃﺮق آﺜﻴﺮة ﻟﺘﻘﺪﻳﺮ اﻟﻤﻌﺎﻟﻢ ﺳﻨﺬآﺮ ﻣﻨﻬﺎ هﻨﺎ ﻓﻘﻂ ﻃﺮﻳﻘﺘﻴﻦ ﺗﺪﺧﻞ ﺿﻤﻦ ﻧﻄﺎق هﺬا اﻟﻤﻘﺮر‬ ‫وهﻤﺎ ﻃﺮﻳﻘﺔ اﻟﻌﺰوم وﻃﺮﻳﻘﺔ اﻟﻤﺮﺑﻌﺎت اﻟﺪﻧﻴﺎ اﻟﺸﺮﻃﻴﺔ‪.‬‬ ‫أوﻻ‪ :‬ﻃﺮﻳﻘﺔ اﻟﻌﺰوم ‪: The Method of Moments‬‬ ‫وﺗﻌﺘﻤﺪ ﻋﻠﻰ ﻣﺴﺎوات ﻋﺰوم اﻟﻌﻴﻨﺔ ﻣﺜﻞ ﻣﺘﻮﺳﻂ اﻟﻌﻴﻨﺔ ‪ z‬واﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ ﻟﻠﻌﻴﻨﺔ ‪ rk‬ﺑﺎﻟﻌﺰوم‬ ‫اﻟﻨﻈﺮﻳﺔ ﻣﺜﻞ اﻟﻤﺘﻮﺳﻂ ‪ µ‬وداﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ‪ ρ k‬وﺣﻞ اﻟﻤﻌﺎدﻻت اﻟﻨﺎﺗﺠﺔ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﻤﻌﺎﻟﻢ اﻟﻤﺮاد‬ ‫ﺗﻘﺪﻳﺮهﺎ‪.‬‬ ‫ﺳﻮف ﻧﺴﺘﻌﺮض اﻟﻄﺮﻳﻘﺔ ﻟﻠﻨﻤﻮذج )‪ AR(p‬آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫‪n‬‬ ‫‪ -١‬ﻳﻘﺪر اﻟﻤﺘﻮﺳﻂ ‪ µ‬ﺑﺎﻟﻤﻘﺪر ‪ z‬اي ‪µˆ = z = ∑i =1 zi n‬‬ ‫‪ -٢‬ﻟﺘﻘﺪﻳﺮ ‪ φ1 ,K ,φ p‬ﻧﺴﺘﺨﺪم اﻟﻌﻼﻗﺔ‪:‬‬ ‫‪٦٧‬‬

‫‪ρ k = φ1 ρ k −1 + φ2 ρ k − 2 + L + φ p ρ k − p , k > 1‬‬

‫واﻟﺘﻲ ﺗﻨﺘﺞ ﻣﻦ ﺿﺮب اﻟﻤﻌﺎدﻟﺔ اﻟﻤﻌﺮﻓﺔ ﻟﻨﻤﻮذج )‪ AR(p‬ﺑﺎﻟﺤﺪ ‪ zt −k − µ‬وأﺧﺬ اﻟﺘﻮﻗﻊ‪ .‬ﻓﻲ‬ ‫اﻟﻤﻌﺎدﻟﺔ اﻟﺴﺎﺑﻘﺔ ﺑﻮﺿﻊ ‪ k = 1, 2,K , p‬ﻧﺤﺼﻞ ﻋﻠﻰ ﻧﻈﺎم اﻟﻤﻌﺎدﻻت اﻟﻤﺴﻤﻰ ﻣﻌﺎدﻻت ﻳﻮل و‬ ‫ووآﺮ ‪ Yule-Walker‬اﻟﺘﺎﻟﻲ‪:‬‬

‫‪ρ1 = φ1 + φ2 ρ1 + L + φ p ρ p −1‬‬

‫‪ρ 2 = φ1 ρ1 + φ2 + L + φ p ρ p −2‬‬ ‫‪M‬‬

‫‪ρ p = φ1 ρ p −1 + φ2 ρ p −2 + L + φ p‬‬

‫و ﺑﺎﻟﺘﻌﻮﻳﺾ ﻋﻦ ‪ ρ k‬ﺑﺎﻟﻤﻘﺪر ‪ rk‬ﻧﺤﺼﻞ ﻋﻠﻰ ﻣﻘﺪرات اﻟﻌﺰوم ﻟﻠﻤﻌﺎﻟﻢ ‪ φˆ1 ,K,φˆp‬آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫ﺑﻮﺿﻊ ﻣﻌﺎدﻻت ﻳﻮل و ووآﺮ ﻋﻠﻰ اﻟﺸﻜﻞ اﻟﻤﺼﻔﻮﻓﻲ‪:‬‬ ‫⎞ ‪rp −1 ⎞ ⎛ φˆ1‬‬ ‫⎟ ⎜‬ ‫⎟ ‪rp −2 ⎟⎟ ⎜ φˆ2‬‬ ‫⎟ ‪M ⎟⎜ M‬‬ ‫⎟ ⎜⎟‬ ‫⎟ ˆ‪1 ⎟⎠ ⎜ φ‬‬ ‫⎠‪⎝ p‬‬

‫‪r2 L rp −2‬‬ ‫‪r1 L rp −3‬‬ ‫‪M‬‬ ‫‪M‬‬ ‫‪M‬‬ ‫‪rp −3 L r1‬‬

‫‪r1‬‬ ‫‪⎛ r1 ⎞ ⎛ 1‬‬ ‫‪⎜r ⎟ ⎜ r‬‬ ‫‪1‬‬ ‫‪⎜ 2⎟=⎜ 1‬‬ ‫‪⎜M⎟ ⎜ M‬‬ ‫‪M‬‬ ‫‪⎜⎜ r ⎟⎟ ⎜⎜ r‬‬ ‫‪⎝ p ⎠ ⎝ p −1 rp −2‬‬

‫وﺑﺤﻞ هﺬﻩ اﻟﻤﻌﺎدﻟﺔ ﻟﻠﻤﻌﺎﻟﻢ‬ ‫⎞ ‪⎛ r1‬‬ ‫⎟ ‪⎜r‬‬ ‫⎟‪⎜ 2‬‬ ‫⎟‪⎜M‬‬ ‫⎟⎟ ⎜⎜‬ ‫⎠ ‪⎝ rp‬‬

‫‪−1‬‬

‫ﺗﻘﺪر ‪ σ 2‬آﺎﻟﺘﺎﻟﻲ‬

‫⎞ ‪rp −1‬‬ ‫⎟⎟ ‪rp −2‬‬ ‫⎟ ‪M‬‬ ‫⎟‬ ‫⎠⎟ ‪1‬‬

‫‪r2 L rp −2‬‬ ‫‪r1 L rp −3‬‬ ‫‪M‬‬ ‫‪M‬‬ ‫‪M‬‬ ‫‪rp −3 L r1‬‬

‫)‬

‫‪r1‬‬ ‫‪1‬‬ ‫‪M‬‬ ‫‪rp −2‬‬

‫‪⎛ φˆ1 ⎞ ⎛ 1‬‬ ‫⎜ ⎟ ⎜‬ ‫‪⎜ φˆ2 ⎟ ⎜ r1‬‬ ‫‪⎜ M ⎟=⎜ M‬‬ ‫⎜ ⎟ ⎜‬ ‫‪⎜ φˆ ⎟ ⎜⎝ rp −1‬‬ ‫⎠‪⎝ p‬‬

‫(‬

‫‪σˆ 2 = γˆ0 1 − φˆ1r1 − φˆ2 r2 −Lφˆp rp‬‬

‫ﺣﻴﺚ‬ ‫‪1‬‬ ‫‪2‬‬ ‫) ‪( zt − z‬‬ ‫∑‬ ‫‪n t =1‬‬ ‫‪n‬‬

‫هﻮ ﺗﺒﺎﻳﻦ اﻟﻌﻴﻨﺔ‪.‬‬ ‫ﺗﻘﺪﻳﺮ اﻟﻌﺰوم ﻟﺒﻌﺾ اﻟﻨﻤﺎذج‪:‬‬ ‫‪ -١‬ﻧﻤﻮذج )‪AR(١‬‬

‫) ‪N ( 0, σ 2‬‬

‫= ‪γˆ0‬‬

‫‪zt − µ = φ1 ( zt −1 − µ ) + at , at‬‬

‫ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ φ1‬هﻮ‬ ‫‪φˆ1 = r1‬‬

‫ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ µ‬هﻮ‬ ‫‪µˆ = z‬‬

‫ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ σ 2‬هﻮ‬

‫)‬

‫(‬

‫‪σˆ 2 = γˆ0 1 − φˆ1r1‬‬

‫ﺣﻴﺚ‬ ‫‪1‬‬ ‫‪2‬‬ ‫) ‪( zt − z‬‬ ‫∑‬ ‫‪n t =1‬‬ ‫‪n‬‬

‫‪٦٨‬‬

‫= ‪γˆ0‬‬

‫‪ -٢‬ﻧﻤﻮذج )‪MA(١‬‬

‫)‬

‫‪2‬‬

‫‪N ( 0, σ‬‬

‫‪zt − µ = at − θ1at −1 , at‬‬

‫ﻹﻳﺠﺎد ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ θ1‬ﻧﺴﺘﺨﺪم اﻟﻌﻼﻗﺔ‬

‫‪−θ1‬‬ ‫‪1 + θ12‬‬

‫= ‪ρ1‬‬

‫وﺑﺘﻌﻮﻳﺾ اﻟﻤﻌﺎﻟﻢ ﺑﻤﻘﺪراﺗﻬﺎ‬ ‫‪−θˆ1‬‬ ‫‪1 + θˆ12‬‬

‫= ‪r1‬‬

‫وﺑﺤﻞ اﻟﻤﻌﺎدﻟﺔ ﻟﻠﻤﻘﺪر ‪ θˆ1‬ﻧﺠﺪ‬ ‫‪−1 ± 1 − 4r1‬‬ ‫‪2r1‬‬

‫= ‪θˆ1‬‬

‫هﺬا اﻟﺤﻞ ﻳﻌﻄﻲ ﻗﻴﻤﺘﻴﻦ ﻟﻠﻤﻘﺪر ‪ θˆ1‬ﻧﺄﺧﺬ اﻟﻘﻴﻤﺔ اﻟﺘﻲ ﺗﺤﻘﻖ ‪ . θˆ1 < 1‬ﻓﻤﺜﻼ ﻟﻮ آﺎﻧﺖ ‪ r1 = −0.4‬ﻓﺈن‬ ‫‪ (θˆ1 ) = −0.77‬و ‪ (θˆ1 ) = 3.27‬وﺑﺎﻟﺘﺎﻟﻲ ﻳﻜﻮن ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ θ1‬هﻮ ‪. θˆ1 = −0.77‬‬ ‫‪1‬‬

‫‪2‬‬

‫‪ -٣‬ﻧﻤﻮذج )‪AR(٢‬‬

‫)‬

‫‪2‬‬

‫‪N ( 0, σ‬‬

‫‪zt − µ = φ1 ( zt −1 − µ ) + φ2 ( zt − 2 − µ ) + at , at‬‬

‫ﺑﺈﺳﺘﺨﺪام ﻣﻌﺎدﻻت ﻳﻮل ووآﺮ ﻣﻘﺪرات اﻟﻌﺰوم ﻟﻠﻤﻌﺎﻟﻢ ‪ φ1‬و ‪ φ2‬هﻲ‬ ‫⎞ ‪⎛ φˆ1 ⎞ ⎛ 1 r1 ⎞ −1 ⎛ r1‬‬ ‫⎜=⎟ ⎜‬ ‫⎠⎟ ‪⎜ φˆ ⎟ ⎝ r1 1 ⎟⎠ ⎜⎝ r2‬‬ ‫⎠‪⎝ 2‬‬

‫وﻣﻨﻬﺎ ﻧﺠﺪ‬ ‫‪r1 − r1r2‬‬ ‫‪r −r‬‬ ‫‪, φˆ2 = 2‬‬ ‫‪2‬‬ ‫‪1 − r1‬‬ ‫‪1− r‬‬

‫‪2‬‬ ‫‪1‬‬ ‫‪2‬‬ ‫‪1‬‬

‫= ‪φˆ1‬‬

‫ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ µ‬هﻮ‬ ‫‪µˆ = z‬‬

‫ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ σ 2‬هﻮ‬

‫)‬

‫(‬

‫‪σˆ 2 = γˆ0 1 − φˆ1r1 − φˆ2 r2‬‬

‫ﺣﻴﺚ‬ ‫‪1‬‬ ‫‪2‬‬ ‫) ‪( zt − z‬‬ ‫∑‬ ‫‪n t =1‬‬ ‫‪n‬‬

‫‪ -٤‬ﻧﻤﻮذج )‪MA(٢‬‬

‫)‬

‫‪2‬‬

‫‪N ( 0, σ‬‬

‫= ‪γˆ0‬‬

‫‪zt − µ = at − θ1at −1 − θ 2 at − 2 , at‬‬

‫ﻹﻳﺠﺎد ﻣﻘﺪرات اﻟﻌﺰوم ﻟﻠﻤﻌﺎﻟﻢ ‪ θ1‬و ‪ θ2‬ﻧﺴﺘﺨﺪم اﻟﻌﻼﻗﺎت‬

‫) ‪−θ1 (1 − θ 2‬‬ ‫‪−θ 2‬‬ ‫= ‪, ρ2‬‬ ‫‪2‬‬ ‫‪2‬‬ ‫‪1 + θ1 + θ 2‬‬ ‫‪1 + θ12 + θ 22‬‬

‫وﺑﺘﻌﻮﻳﺾ اﻟﻤﻘﺪرات ‪ r1‬و ‪ r2‬ﻧﺤﺼﻞ ﻋﻠﻰ ﻣﻘﺪرات اﻟﻌﺰوم ﻟﻠﻤﻌﺎﻟﻢ ‪ θ1‬و ‪θ2‬‬

‫)‬

‫(‬

‫= ‪ρ1‬‬

‫‪−θˆ1 1 − θˆ2‬‬ ‫‪−θˆ2‬‬ ‫‪,‬‬ ‫‪r‬‬ ‫=‬ ‫‪2‬‬ ‫‪1 + θˆ12 + θˆ22‬‬ ‫‪1 + θˆ12 + θˆ22‬‬

‫‪٦٩‬‬

‫= ‪r1‬‬

‫وﻧﺤﻞ ﻟﻜﻞ ﻣﻦ ‪ θˆ1‬و ‪ θˆ2‬وﻧﺄﺧﺬ اﻟﺤﻠﻮل اﻟﺘﻲ ﺗﺤﻘﻖ ‪. θ 2 − θ1 < 1, θ 2 + θ1 < 1, θ 2 < 1‬‬ ‫‪ -٥‬ﻧﻤﻮذج )‪ARMA(١،١‬‬

‫)‬

‫‪2‬‬

‫‪N ( 0, σ‬‬

‫‪zt − µ = φ1 ( zt −1 − µ ) + at − θ1at −1 , at‬‬

‫ﻹﻳﺠﺎد ﻣﻘﺪرات اﻟﻌﺰوم ﻟﻠﻤﻌﺎﻟﻢ ‪ θ1‬و ‪ φ1‬ﻧﺴﺘﺨﺪم اﻟﻌﻼﻗﺎت‬

‫‪(1 − φ1θ1 )(φ1 − θ1 ) , ρ = (1 − φ1θ1 )(φ1 − θ1 ) φ‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫‪2‬‬ ‫‪2‬‬ ‫‪1 + θ1 − 2φ1θ1‬‬

‫‪1 + θ1 − 2φ1θ1‬‬

‫= ‪ρ1‬‬

‫وﺑﺘﻌﻮﻳﺾ اﻟﻤﻘﺪرات ‪ r1‬و ‪ r2‬ﻧﺤﺼﻞ ﻋﻠﻰ ﻣﻘﺪرات اﻟﻌﺰوم ﻟﻠﻤﻌﺎﻟﻢ ‪ θ1‬و ‪φ1‬‬ ‫) ˆ‪1 − φˆ θˆ )(φˆ − θ‬‬ ‫(‬ ‫ˆ‪φ‬‬ ‫= ‪r‬‬ ‫‪1‬‬

‫‪1 1‬‬

‫‪1‬‬

‫ˆ‪1 + θˆ 2 − 2φˆ θ‬‬

‫‪1‬‬

‫‪1 1‬‬

‫‪2‬‬

‫‪1‬‬

‫) ˆ‪1 − φˆ θˆ )(φˆ − θ‬‬ ‫(‬ ‫= ‪r‬‬ ‫‪,‬‬ ‫‪1‬‬

‫‪1‬‬

‫‪1 1‬‬

‫ˆ‪1 + θˆ 2 − 2φˆ θ‬‬

‫‪1 1‬‬

‫‪1‬‬

‫‪1‬‬

‫وﺑﻘﺴﻤﺔ اﻟﻤﻌﺎدﻟﺔ اﻟﻤﻌﺮﻓﺔ ﻟﻠﻤﻘﺪر ‪ r2‬ﻋﻠﻰ اﻟﻤﻌﺎدﻟﺔ اﻟﻤﻌﺮﻓﺔ ﻟﻠﻤﻘﺪر ‪ r1‬ﻧﺠﺪ‬ ‫‪r2‬‬ ‫‪r1‬‬

‫= ‪φˆ1‬‬

‫وهﻮ ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ . φ1‬ﻹﻳﺠﺎد ﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪ θ1‬ﻧﻌﻮض ﻋﻦ ‪ φˆ1‬ﻓﻲ اﻟﻤﻌﺎدﻟﺔ اﻟﻤﻌﺮﻓﺔ‬ ‫ﻟﻠﻤﻘﺪر ‪ r1‬ﻧﺠﺪ‬ ‫⎞ ˆ ‪⎛ r2 ˆ ⎞ ⎛ r2‬‬ ‫⎟ ‪⎜ 1 − r θ1 ⎟ ⎜ r − θ1‬‬ ‫‪1‬‬ ‫‪⎠⎝ 1‬‬ ‫⎠‬ ‫⎝ = ‪r1‬‬ ‫‪r‬‬ ‫‪1 + θˆ12 − 2 2 θˆ1‬‬ ‫‪r1‬‬ ‫وﻧﺤﻞ اﻟﻤﻌﺎدﻟﺔ اﻟﺘﺮﺑﻴﻌﻴﺔ اﻟﻨﺎﺗﺠﺔ ﻟﻠﻤﻘﺪر ‪ θˆ1‬و ﻧﺄﺧﺬ اﻟﻘﻴﻤﺔ اﻟﺘﻲ ﺗﺤﻘﻖ ‪. θˆ1 < 1‬‬

‫ﺗﻤﺎرﻳﻦ‪ :‬أوﺟﺪ ﻣﻘﺪرات اﻟﻌﺰوم ﻟﻤﻌﺎﻟﻢ اﻟﻨﻤﺎذج اﻟﺘﺎﻟﻴﺔ‬ ‫‪ARIMA(١،١،١), ARIMA(٢،١،٠), ARIMA(٠،١،٢), ARIMA(١،٢،٠),‬‬ ‫‪ARIMA(٠،٢،١), ARIMA(٠،٢،٠).‬‬ ‫ﻣﻼﺣﻈﺔ‪ :‬ﻣﻘﺪرات اﻟﻌﺰوم ﺗﺴﺘﺨﺪم آﻘﻴﻢ أوﻟﻴﺔ ﻹﻳﺠﺎد ﻣﻘﺪرات أآﺜﺮ دﻗﺔ‪.‬‬

‫ﺛﺎﻧﻴﺎ‪ :‬ﻃﺮﻳﻘﺔ اﻟﻤﺮﺑﻌﺎت اﻟﺪﻧﻴﺎ اﻟﺸﺮﻃﻴﺔ ‪Conditional Least Square‬‬ ‫‪: Method‬‬ ‫ﻟﻨﻤﺎذج )‪ ARMA(p,q‬واﻟﺘﻲ ﺗﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫)‬

‫‪2‬‬

‫‪N ( 0, σ‬‬

‫‪φ p ( B )( zt − µ ) = θ q ( B ) at , at‬‬

‫ﺣﻴﺚ ) ‪ φ p ( B‬و ) ‪ θ q ( B‬ﻻﻳﻮﺟﺪ ﺑﻴﻨﻬﺎ ﺟﺬور ﻣﺸﺘﺮآﺔ وﺟﺬور اﻟﻤﻌﺎدﻟﺔ ‪ θ q ( B ) = 0‬ﺗﻘﻊ ﺟﻤﻴﻌﻬﺎ‬ ‫ﺧﺎرج داﺋﺮة اﻟﻮﺣﺪة ) ﺷﺮط اﻹﻧﻘﻼب(‪ .‬ﺑﺈﻋﺎدة آﺘﺎﺑﺔ اﻟﻨﻤﻮذج اﻟﺴﺎﺑﻖ ﻟﻸﺧﻄﺎء ‪ at‬آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫)‪φp (B‬‬ ‫)‪(z − µ‬‬ ‫‪θq ( B ) t‬‬

‫= ‪at‬‬

‫اﻟﻄﺮف اﻷﻳﻤﻦ ﻳﻤﻜﻦ إﻋﺘﺒﺎرة آﺪاﻟﺔ ﻓﻲ اﻟﻤﻌﺎﻟﻢ } ‪ φ = {φ1 , φ2 ,K , φ p‬و } ‪ θ = {θ1 ,θ 2 ,K ,θ q‬و ‪µ‬‬

‫و ﻳﻜﺘﺐ‬

‫)‪) (z − µ‬‬ ‫)‬ ‫‪t‬‬

‫‪٧٠‬‬

‫‪p‬‬

‫‪−L − φ p B‬‬

‫‪2‬‬

‫‪p‬‬

‫‪−L −θ pB‬‬

‫‪2‬‬

‫‪(1 − φ B − φ B‬‬ ‫‪(1 − θ B − θ B‬‬ ‫‪2‬‬

‫‪2‬‬

‫‪1‬‬

‫‪1‬‬

‫= ) ‪at ( φ, θ, µ‬‬

‫ﺗﻌﺘﻤﺪ ﻃﺮﻳﻘﺔ اﻟﻤﺮﺑﻌﺎت اﻟﺪﻧﻴﺎ اﻟﺸﺮﻃﻴﺔ و ﻟﻤﺸﺎهﺪات ﻣﻌﻄﺎة } ‪ z = {z1 , z2 ,K, zn‬ﻋﻠﻰ ﺗﺼﻐﻴﺮ‬ ‫اﻟﺪاﻟﺔ‬ ‫‪n‬‬

‫) ‪∑ a ( φ, θ, µ z‬‬ ‫‪2‬‬ ‫‪t‬‬

‫= ) ‪min Sc ( φ, θ, µ‬‬ ‫‪φ,θ , µ‬‬

‫‪t = p +1‬‬

‫وﺣﻞ اﻟﻤﻌﺎدﻻت اﻟﻄﺒﻴﻌﻴﺔ ‪ Normal Equations‬اﻟﻨﺎﺗﺠﺔ اﻟﺘﺎﻟﻴﺔ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﻤﻘﺪرات‪.‬‬ ‫∂‬ ‫‪∂ n 2‬‬ ‫) ‪Sc ( φ, θ, µ‬‬ ‫=‬ ‫‪∑ at ( φ, θ, µ z ) φ=φˆ = 0‬‬ ‫ˆ‪φ= φ‬‬ ‫‪∂φ‬‬ ‫∂‬ ‫‪φ‬‬ ‫‪t = p +1‬‬ ‫ˆ‬ ‫‪θ =θ‬‬ ‫ˆ‪θ =θ‬‬ ‫ˆ‪µ = µ‬‬

‫ˆ‪µ = µ‬‬

‫∂‬ ‫‪∂ n 2‬‬ ‫) ‪Sc ( φ, θ, µ‬‬ ‫=‬ ‫‪∑p+1 at ( φ, θ, µ z ) φ=φˆ = 0‬‬ ‫ˆ‪φ= φ‬‬ ‫‪θ‬‬ ‫‪∂θ‬‬ ‫∂‬ ‫‪t‬‬ ‫=‬ ‫ˆ‪θ = θ‬‬ ‫ˆ‪θ =θ‬‬ ‫ˆ‪µ = µ‬‬

‫ˆ‪µ = µ‬‬

‫∂‬ ‫‪∂ n 2‬‬ ‫) ‪Sc ( φ, θ, µ‬‬ ‫=‬ ‫‪∑ at ( φ, θ, µ z ) φ=φˆ = 0‬‬ ‫ˆ‪φ= φ‬‬ ‫‪∂µ‬‬ ‫‪∂µ t = p +1‬‬ ‫ˆ‪θ = θ‬‬ ‫ˆ‪θ =θ‬‬ ‫ˆ‪µ = µ‬‬

‫ˆ‪µ = µ‬‬

‫هﺬﻩ اﻟﻤﻘﺪرات ﺗﺴﻤﻰ ﺷﺮﻃﻴﺔ ﻷﻧﻨﺎ هﻨﺎ ﻧﺸﺘﺮط ان اﻟﻘﻴﻢ ‪ a p = a p −1 = L = a p +1− q = 0‬أي ﻣﺴﺎوﻳﺔ‬ ‫ﻟﺘﻮﻗﻌﻬﺎ‪ ) .‬ﻻﺣﻆ أن اﻟﺘﺠﻤﻴﻊ ﻓﻲ اﻟﻤﻌﺎدﻻت اﻟﺴﺎﺑﻘﺔ ﻳﺒﺪأ ﻣﻦ اﻟﻘﻴﻤﺔ ‪.( t = p + 1‬‬ ‫ﻳﻘﺪر اﻟﺘﺒﺎﻳﻦ ‪ σ 2‬ﻣﻦ‬

‫)‬

‫(‬

‫‪Sc φˆ , θˆ , µ‬‬

‫)‪n − ( p + q + 1‬‬

‫= ‪σˆ 2‬‬

‫ﺗﻘﺪﻳﺮات اﻟﻤﺮﺑﻌﺎت اﻟﺪﻧﻴﺎ اﻟﺸﺮﻃﻴﺔ ﻟﺒﻌﺾ اﻟﻨﻤﺎذج‪:‬‬ ‫‪ -١‬ﻧﻤﻮذج )‪AR(١‬‬

‫)‬

‫ﻟﺘﺒﺴﻴﻂ اﻹﺷﺘﻘﺎﻗﺎت ﺳﻮف ﻧﺴﺘﺒﺪل ‪ µ‬ﺑﻤﻘﺪرهﺎ ‪z‬‬

‫‪2‬‬

‫)‬

‫‪N ( 0, σ‬‬

‫‪zt − µ = φ1 ( zt −1 − µ ) + at , at‬‬

‫‪N ( 0, σ‬‬

‫‪zt − z = φ1 ( zt −1 − z ) + at , at‬‬

‫‪2‬‬

‫ﻟﻤﺸﺎهﺪات ﻣﻌﻄﺎة } ‪ z = {z1 , z2 ,K, zn‬ﻧﻜﺘﺐ اﻷﺧﻄﺎء‬

‫‪at (φ1 ) = ( zt − z ) − φ1 ( zt −1 − z ) , t = 2,3,L, n‬‬

‫وﺗﺮﺑﻴﻊ اﻟﻄﺮﻓﻴﻦ واﻟﺠﻤﻊ ﻋﻠﻰ آﻞ اﻟﻤﺸﺎهﺪات‬ ‫‪a (φ1 ) = ⎡⎣( zt − z ) − φ1 ( zt −1 − z )⎤⎦ , t = 2,3,L , n‬‬ ‫‪2‬‬

‫‪2‬‬

‫‪2‬‬ ‫‪t‬‬

‫‪n‬‬

‫‪n‬‬

‫‪t =2‬‬

‫‪t =2‬‬

‫⎦⎤) ‪Sc (φ1 ) = ∑ at2 (φ1 ) = ∑ ⎡⎣( zt − z ) − φ1 ( zt −1 − z‬‬

‫وهﺬﻩ داﻟﺔ ﻟﻠﻤﻌﻠﻢ ‪ φ1‬ﻓﻘﻂ ‪ ،‬ﻧﺸﺘﻖ اﻟﻤﻌﺎدﻟﺔ اﻟﺴﺎﺑﻘﺔ ﺑﺎﻟﻨﺴﺒﺔ ﻟﻠﻤﻌﻠﻢ ‪ φ1‬وﺗﻜﻮن اﻟﻨﺘﻴﺠﺔ ﻣﺴﺎوﻳﺔ ﻟﻠﺼﻔﺮ‬ ‫ﻋﻨﺪﻣﺎ ‪ φ1 = φˆ1‬أي‬

‫‪٧١‬‬

‫‪2‬‬

‫‪n‬‬

‫‪n‬‬

‫⎦⎤) ‪Sc (φ1 ) = ∑ at2 (φ1 ) = ∑ ⎡⎣( zt − z ) − φ1 ( zt −1 − z‬‬ ‫‪t =2‬‬

‫‪t =2‬‬

‫‪2‬‬ ‫∂‬ ‫∂‬ ‫= ) ‪Sc (φ1‬‬ ‫⎦⎤) ‪⎡⎣( zt − z ) − φ1 ( zt −1 − z‬‬ ‫∑‬ ‫‪∂φ1‬‬ ‫‪∂φ1 t =2‬‬ ‫‪n‬‬

‫‪n‬‬

‫⎦⎤) ‪= ∑ −2 ( zt −1 − z ) ⎡⎣ ( zt − z ) − φ1 ( zt −1 − z‬‬ ‫‪t= 2‬‬

‫‪n‬‬ ‫∂‬ ‫‪Sc (φ1 ) ˆ = ∑ −2 ( zt −1 − z ) ⎡⎣( zt − z ) − φˆ1 ( zt −1 − z )⎦⎤ = 0‬‬ ‫‪φ1 =φ1‬‬ ‫‪∂φ1‬‬ ‫‪t =2‬‬ ‫‪n‬‬

‫‪∴ ∑ ( zt −1 − z ) ⎡⎣( zt − z ) − φˆ1 ( zt −1 − z )⎤⎦ = 0‬‬ ‫‪t =2‬‬

‫‪n‬‬

‫‪n‬‬

‫‪t =2‬‬

‫‪t= 2‬‬

‫‪2‬‬ ‫‪∑ ( zt −1 − z )( zt − z ) − φˆ1 ∑ ( zt −1 − z ) = 0‬‬

‫أي‬ ‫) ‪− z )( zt − z‬‬

‫‪n‬‬

‫‪t −1‬‬

‫‪∑( z‬‬ ‫‪t =2‬‬

‫‪2‬‬

‫)‪−z‬‬

‫‪n‬‬

‫‪t −1‬‬

‫‪∑( z‬‬

‫= ‪φˆ1‬‬

‫‪t =2‬‬

‫وهﻮ ﻣﻘﺪر اﻟﻤﺮﺑﻌﺎت اﻟﺪﻧﻴﺎ اﻟﺸﺮﻃﻴﺔ ﻟﻠﻤﻌﻠﻢ ‪. φ1‬‬ ‫ﺗﻤﺮﻳﻦ‪ :‬ﻗﺎرن ﺑﻴﻦ هﺬا اﻟﻤﻘﺪر وﻣﻘﺪر اﻟﻌﺰوم ﻟﻠﻤﻌﻠﻢ ‪. φ1‬‬ ‫‪ -٢‬ﻧﻤﻮذج )‪MA(١‬‬

‫) ‪N ( 0,σ 2‬‬

‫‪zt − µ = at − θ1at −1 , at‬‬

‫ﻟﺘﺒﺴﻴﻂ اﻹﺷﺘﻘﺎﻗﺎت ﺳﻮف ﻧﺴﺘﺒﺪل ‪ µ‬ﺑﻤﻘﺪرهﺎ ‪ z‬وﻧﻌﻤﻞ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﻌﺪﻟﺔ ﻟﻠﻤﺘﻮﺳﻂ‬ ‫‪ xt = zt − z‬ﻓﻴﺼﺒﺢ اﻟﻨﻤﻮذج‬ ‫) ‪N ( 0,σ 2‬‬

‫وﺑﻜﺘﺎﺑﺔ اﻟﻤﻌﺎدﻟﺔ اﻷﺧﻴﺮة ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫‪xt = at − θ1at −1 , at‬‬ ‫‪at = xt − θ1at −1‬‬

‫وﻟﻤﺸﺎهﺪات ﻣﻌﻄﺎة ‪ x1 , x2 ,K , xn‬و ﺑﻮﺿﻊ ‪ a0 = 0‬ﺷﺮﻃﻴﺎ ﻧﻜﺘﺐ اﻷﺧﻄﺎء‬

‫‪a1 = x1‬‬ ‫‪a2 = x2 − θ1a1‬‬ ‫‪a3 = x3 − θ1a2‬‬ ‫‪M‬‬ ‫‪an = xn − θ1an −1‬‬

‫وﺑﺎﻟﺘﺎﻟﻲ‬ ‫‪n‬‬

‫‪Sc (θ1 ) = ∑ at2‬‬ ‫‪t =1‬‬

‫اﻟﺪاﻟﺔ اﻟﺴﺎﺑﻘﺔ ﻏﻴﺮ ﺧﻄﻴﺔ ﻓﻲ اﻟﻤﻌﻠﻢ ‪ θ1‬و ﻳﻤﻜﻦ إﻳﺠﺎد ﻗﻴﻤﺔ ‪ θ1‬واﻟﺘﻲ ﺗﺼﻐﺮ ) ‪ Sc (θ1‬ﺑﻄﺮق اﻟﺒﺤﺚ‬ ‫اﻟﻌﺪدﻳﺔ ﻣﺜﻞ اﻟﺒﺤﺚ اﻟﺸﺒﻜﻲ ﻓﻲ اﻟﻤﺠﺎل )‪ (-١،١‬أو إﺳﺘﺨﺪام ﻃﺮﻳﻘﺔ ﺟﺎوس‪-‬ﻧﻴﻮﺗﻦ واﻟﺘﻲ ﺗﺘﻠﺨﺺ‬ ‫ﻓﻲ ﺗﻘﺮﻳﺐ ) ‪ at = at (θ1‬ﺑﺪاﻟﺔ ﺧﻄﻴﺔ ﻟﻠﻤﻌﻠﻢ ‪ θ1‬ﺣﻮل ﻗﻴﻤﺔ أوﻟﻴﺔ * ‪ θ‬ﻣﺜﻼ أي‬

‫‪٧٢‬‬

‫) * ‪dat (θ‬‬

‫اﻟﻤﺸﺘﻘﺔ‬

‫) * ‪dat (θ‬‬ ‫‪d θ1‬‬

‫‪dθ1‬‬

‫) * ‪at (θ1 ) ≈ at (θ * ) + (θ1 − θ‬‬

‫ﻳﻤﻜﻦ ﺣﺴﺎﺑﻬﺎ ﺗﻜﺮارﻳﺎ وذﻟﻚ ﺑﺈﺷﺘﻘﺎق ﻃﺮﻓﻲ اﻟﻤﻌﺎدﻟﺔ ‪ at = xt − θ1at −1‬ﺑﺎﻟﻨﺴﺒﺔ‬

‫ﻟﻠﻤﻌﻠﻢ ‪ θ1‬ﻟﻨﺤﺼﻞ ﻋﻠﻰ‬

‫) ‪da0 (θ1‬‬ ‫و ﺑﻘﻴﻤﺔ أوﻟﻴﺔ ‪= 0‬‬ ‫‪dθ1‬‬

‫) ‪dat (θ1 ) θ1dat −1 (θ1‬‬ ‫=‬ ‫) ‪+ at −1 (θ1‬‬ ‫‪dθ1‬‬ ‫‪dθ1‬‬

‫‪ .‬اﻟﻤﻌﺎدﻟﺔ‬ ‫) * ‪dat (θ‬‬ ‫‪dθ1‬‬

‫)‬

‫*‬

‫‪−θ‬‬

‫‪1‬‬

‫‪) + (θ‬‬

‫*‬

‫‪at (θ1 ) ≈ at (θ‬‬

‫ﺧﻄﻴﺔ ﻓﻲ اﻟﻤﻌﻠﻢ ‪ θ1‬وﺑﺎﻟﺘﺎﻟﻲ ﺑﺎﻻﻣﻜﺎن ﺗﺼﻐﻴﺮ ﻣﺠﻤﻮع اﻟﻤﺮﺑﻌﺎت‬ ‫‪n‬‬

‫‪Sc (θ1 ) = ∑ at2‬‬ ‫‪t=1‬‬

‫ﺗﺤﻠﻴﻠﻴﺎ ﻟﻨﺤﺼﻞ ﻋﻠﻲ ﻣﻘﺪر ﺟﺪﻳﺪ وأﻓﻀﻞ ﻟﻠﻤﻌﻠﻢ ‪ θ1‬وﻧﻜﺮر هﺬﻩ اﻟﻌﻤﻠﻴﺔ ﺑﺈﺳﺘﺒﺪال * ‪ θ‬ﺑﺎﻟﻤﻘﺪر‬ ‫اﻟﺠﺪﻳﺪ وﻧﺴﺘﻤﺮ ﺣﺘﻰ ﻳﺼﺒﺢ اﻟﻔﺮق ﺑﻴﻦ ﻣﻘﺪرﻳﻦ ﺗﺎﻟﻴﻴﻦ ﺻﻐﻴﺮ ﺟﺪا أو اﻟﻨﻘﺺ ﻓﻲ ﻣﺠﻤﻮع اﻟﻤﺮﺑﻌﺎت‬ ‫ﺻﻐﻴﺮ ﺟﺪا‪ .‬ﻣﻤﻜﻦ إﺳﺘﺨﺪام ﻃﺮﻳﻘﺔ اﻟﻌﺰوم ﻹﻳﺠﺎد اﻟﻘﻴﻤﺔ أﻷوﻟﻴﺔ * ‪ θ‬ﻟﻜﻲ ﻧﺤﺼﻞ ﻋﻠﻰ ﺗﻘﺎرب‬ ‫ﺳﺮﻳﻊ‪ .‬ﻃﺒﻌﺎ اﻟﻄﺮﻳﻘﺔ اﻟﺴﺎﺑﻘﺔ ﻻﺗﺘﻢ ﻳﺪوﻳﺎ ﺑﻞ ﺗﺤﺘﺎج إﻟﻰ ﺣﺎﺳﺐ ﻟﺬﻟﻚ‪.‬‬ ‫ﻳﻼﺣﻆ أن ﺗﻘﺪﻳﺮ اﻟﻤﻌﺎﻟﻢ ﻟﻠﻨﻤﻮذج ﻓﻲ ﺣﺎﻟﺔ ﻧﻤﺎذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك او اﻟﻨﻤﺎذج اﻟﻤﺨﺘﻠﻄﺔ اﻟﺘﻲ‬ ‫ﺗﺤﻮي ﻋﻠﻰ ﻣﺘﻮﺳﻂ ﻣﺘﺤﺮك ﺗﺸﻜﻞ ﺗﻌﻘﻴﺪا ﻷﻧﻬﺎ ﺗﺤﻮى ﻣﻌﺎﻟﻢ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﺑﺸﻜﻞ ﻏﻴﺮ ﺧﻄﻲ‬ ‫وﻟﻬﺬا ﺗﺤﺘﺎج إﻟﻰ ﻃﺮق ﻋﺪدﻳﺔ ﻟﺤﻠﻬﺎ آﻤﺎ ﺷﺎهﺪﻧﺎ ﻓﻲ ﺣﺎﻟﺔ اﻟﻨﻤﻮذج )‪ MA(١‬وهﻮ أﺑﺴﻄﻬﺎ ﺟﻤﻴﻌﺎ‪.‬‬ ‫ﺳﻮف ﻧﻜﺘﻔﻲ ﻓﻲ ﻣﻘﺮرﻧﺎ هﺬا ﻋﻠﻰ اﻟﻄﺮﻳﻘﺘﻴﻦ اﻟﺴﺎﺑﻘﺔ وﻟﻜﻦ ﻧﺬآﺮ ﺑﻌﺾ اﻟﻄﺮق اﻻﺧﺮى اﻟﻤﺴﺘﺨﺪﻣﺔ‬ ‫ﻓﻲ ﺗﻘﺪﻳﺮ ﻣﻌﺎﻟﻢ اﻟﻨﻤﻮذج ﻣﺜﻞ‪:‬‬ ‫‪ -١‬ﻃﺮﻳﻘﺔ اﻷرﺟﺤﻴﺔ اﻟﻌﻈﻤﻰ ‪Maximum Likelihood Method‬‬ ‫‪ -٢‬ﻃﺮﻳﻘﺔ اﻟﻤﺮﺑﻌﺎت اﻟﺪﻧﻴﺎ ﻏﻴﺮ اﻟﺸﺮﻃﻴﺔ ‪Unconditional Least Squares Method‬‬ ‫‪ -٣‬ﻃﺮق اﻟﺘﻘﺪﻳﺮﻏﻴﺮ اﻟﺨﻄﻴﺔ ‪Nonlinear Estimation Methods‬‬ ‫ﺗﺸﺨﻴﺺ وإﺧﺘﺒﺎر اﻟﻨﻤﻮذج ‪: Model Checking and Diagnostics‬‬ ‫ﺑﻌﺪ اﻟﺘﻌﺮف ﻋﻠﻰ ﻧﻤﻮذج ﻣﺒﺪﺋﻲ وﺗﻘﺪﻳﺮ ﻣﻌﺎﻟﻢ هﺬا اﻟﻨﻤﻮذج ﻧﺠﺮي ﺑﻌﺾ اﻟﺘﺸﺨﻴﺼﺎت ﻋﻠﻰ اﻟﺒﻮاﻗﻲ‬ ‫أو أﺧﻄﺎء اﻟﺘﻄﺒﻴﻖ )اﻧﻈﺮ ﺗﻌﺮﻳﻒ ‪ (٤‬ﻟﻨﺮى ﻣﺪى ﻣﻄﺎﺑﻘﺔ اﻟﻨﻤﻮذج ﻟﻠﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﺸﺎهﺪة ‪ ،‬وﻳﻔﺘﺮض‬ ‫أن اﻟﺒﻮاﻗﻲ هﻲ ﻣﻘﺪرات ﻟﻤﺘﺴﻠﺴﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء ‪ at‬واﻟﺘﻲ ﻧﻔﺘﺮض اﻧﻬﺎ ﻣﻮزﻋﺔ ﻃﺒﻴﻌﻴﺎ ﺑﻤﺘﻮﺳﻂ‬ ‫ﺻﻔﺮي وﺗﺒﺎﻳﻦ ‪ . σ 2‬اﻟﺒﻮاﻗﻲ ﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬ ‫‪et = zt − zˆt = aˆt , t = 1, 2,..., n‬‬

‫أي ان اﻟﺒﻮاﻗﻲ هﻲ اﻟﻘﻴﻢ اﻟﻤﺸﺎهﺪة ﻧﺎﻗﺺ اﻟﻘﻴﻢ اﻟﻤﻄﺒﻘﺔ‪.‬‬ ‫ﻳﻘﻮم اﻟﺘﺸﺨﻴﺺ واﻹﺧﺘﺒﺎرات ﻋﻠﻰ ﻓﺤﺺ اﻟﺒﻮاﻗﻲ واﻟﻨﻈﺮ ﻓﻲ ﻣﺪى ﺗﺤﻘﻴﻘﻬﺎ ﻟﻔﺮﺿﻴﺎت اﻟﻨﻤﻮذج‬ ‫واﻟﺘﻲ هﻲ‪:‬‬ ‫‪ -١‬ﻣﺘﻮﺳﻂ ﺻﻔﺮي‬ ‫‪ -٢‬اﻟﻌﺸﻮاﺋﻴﺔ‬ ‫‪٧٣‬‬

‫‪ -٣‬ﻋﺪم اﻟﺘﺮاﺑﻂ‬ ‫‪ -٤‬ﻣﻮزﻋﺔ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ )ﻣﺴﺘﻘﻞ وﻣﺘﻄﺎﺑﻖ ﺑﻤﺘﻮﺳﻂ ﺻﻔﺮي وﺗﺒﺎﻳﻦ ‪σ‬‬ ‫أي ) ‪( at IIDN ( 0, σ 2‬‬ ‫‪2‬‬

‫ﻟﻬﺬا ﻓﺈﻧﻨﺎ ﻧﺠﺮي ﺗﺸﺨﻴﺺ وهﻮ ﻣﺠﻤﻮﻋﺔ ﻣﻦ اﻹﺧﺘﺒﺎرات ﻋﻠﻲ اﻟﺒﻮاﻗﻲ ﻟﻨﺮي ﻓﻴﻤﺎ إذا آﺎﻧﺖ ﺗﺤﻘﻖ‬ ‫هﺬﻩ اﻟﺸﺮوط وﻓﻲ هﺬﻩ اﻟﺤﺎﻟﺔ ﻧﻌﺘﺒﺮ اﻟﻨﻤﻮذج اﻟﻤﻄﺒﻖ ﻣﻘﺒﻮﻻ أﻣﺎ إذا ﻓﺸﻞ اﺣﺪ هﺬﻩ اﻹﺧﺘﺒﺎرات‬ ‫ﻓﻴﺠﺐ ﻋﻠﻴﻨﺎ إﻋﺎدة اﻟﻨﻈﺮ وإﻗﺘﺮاح ﻧﻤﻮذج ﺁﺧﺮ‬ ‫أوﻻ‪ :‬إﺧﺘﺒﺎر اﻟﻤﺘﻮﺳﻂ‪:‬‬ ‫‪H 0 : E ( at ) = 0‬‬ ‫‪H 1 : E ( at ) ≠ 0‬‬

‫وهﻮ إﺧﺘﺒﺎر ﺑﺬﻳﻠﻴﻦ وﻧﺴﺘﺨﺪم ﻓﻴﺔ اﻹﺣﺼﺎﺋﺔ‬

‫‪e‬‬

‫) ‪se ( e‬‬

‫= ‪ u‬واﻟﺘﻲ ﻟﻬﺎ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ ﻗﻴﺎﺳﻲ ﻓﻌﻨﺪ‬

‫ﻣﺴﺘﻮى ﻣﻌﻨﻮﻳﺔ ‪ α = 0.05‬ﻧﻌﺘﺒﺮ ان ‪ E ( at ) = 0‬إذا آﺎﻧﺖ ‪ ) u < 1.96‬هﺬا ﻋﻠﻲ إﻋﺘﺒﺎر ان‬ ‫ﺣﺠﻢ اﻟﻌﻴﻨﺔ اآﺒﺮ ﻣﻦ ‪ ٣٠‬وﺣﺪة وهﺬا داﺋﻤﺎ ﻣﺘﺤﻘﻖ ﻟﻠﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﺘﻲ ﻧﺪرﺳﻬﺎ (‬ ‫ﺛﺎﻧﻴﺎ‪ :‬إﺧﺘﺒﺎر اﻟﻌﺸﻮاﺋﻴﺔ‪:‬‬ ‫ﻧﺨﺘﺒﺮ ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ ﺑﻮاﺳﻄﺔ إﺧﺘﺒﺎر اﻟﺠﺮي ‪ Runs test‬ﺣﻮل اﻟﻤﺘﻮﺳﻂ وﺣﻮل اﻟﺼﻔﺮ‬ ‫وهﻮ اﺣﺪ اﻹﺧﺘﺒﺎرات اﻟﻼﻣﻌﻠﻤﻴﺔ ) ﻳﻮﺟﺪ آﺜﻴﺮ ﻣﻦ اﻹﺧﺘﺒﺎرات ﻟﻠﻌﺸﻮاﺋﻴﺔ ﻳﺪرﺳﻬﺎ اﻟﻄﺎﻟﺐ ﻓﻲ‬ ‫اﻟﻤﻘﺮر ‪ ٢٤١‬ﺑﺤﺚ وﻟﻜﻦ ﻧﻜﺘﻔﻲ هﻨﺎ ﺑﻬﺬا اﻹﺧﺘﺒﺎر(‪.‬‬ ‫ﺛﺎﻟﺜﺎ‪ :‬إﺧﺘﺒﺎر اﻟﺘﺮاﺑﻂ أو اﻹﺳﺘﻘﻼل‪:‬‬ ‫ﻳﺨﺘﺒﺮ ﺗﺮاﺑﻂ أو إﺳﺘﻘﻼل اﻟﺒﻮاﻗﻲ ﺑﻮاﺳﻄﺔ إﺧﺘﺒﺎر اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ‪Autocorrelation‬‬ ‫‪ test‬وذﻟﻚ ﺑﺤﺴﺎب ورﺳﻢ اﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﻌﻴﻨﻴﺔ ‪ SACF‬ﻟﻠﺒﻮاﻗﻲ وﻣﻘﺎرﻧﺘﻬﺎ ﻣﻊ داﻟﺔ‬ ‫اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻤﺘﺴﻠﺴﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‪.‬‬ ‫اﻹﺧﺘﺒﺎر‬ ‫‪H 0 : ρ1 = 0‬‬

‫‪H1 : ρ1 ≠ 0‬‬ ‫‪r1‬‬ ‫ﺣﻴﺚ اﻹﺣﺼﺎﺋﺔ‬ ‫) ‪se ( r1‬‬

‫= ‪ u‬ﻟﻬﺎ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ ﻗﻴﺎﺳﻲ ﻓﻌﻨﺪ ﻣﺴﺘﻮى ﻣﻌﻨﻮﻳﺔ ‪ α = 0.05‬ﻧﻌﺘﺒﺮ‬

‫ان ‪ ρ1 = 0‬إذا آﺎﻧﺖ ‪. u < 1.96‬‬ ‫راﺑﻌﺎ‪ :‬إﺧﺘﺒﺎر ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‪:‬‬ ‫ﻧﺨﺘﺒﺮ ﻓﻲ ﻣﺎ إذا آﺎﻧﺖ اﻟﺒﻮاﻗﻲ ﻣﻮزﻋﺔ ﻃﺒﻴﻌﻴﺎ وذﻟﻚ ﺑﻌﺪة ﻃﺮق ﻣﺜﻞ‪:‬‬ ‫‪ -١‬إﺧﺘﺒﺎر ﺣﺴﻦ اﻟﺘﻄﺎﺑﻖ ‪ Goodness of Fit Test‬وﻧﺴﺘﺨﺪم اﻹﺧﺘﺒﺎر اﻟﻼﻣﻌﻠﻤﻲ‬ ‫آﻮﻟﻤﻮﺟﻮروف‪ -‬ﺳﻤﻴﺮﻧﻮف ‪. Kolmogorov-Smirnov Test‬‬ ‫‪ -٢‬ﻣﺨﻄﻂ اﻹﺣﺘﻤﺎل اﻟﻄﺒﻴﻌﻲ ‪. Normal Probability Plot‬‬ ‫‪ -٣‬ﻣﺨﻄﻂ اﻟﺮﺑﻴﻌﺎت‪-‬اﻟﺮﺑﻴﻌﺎت ‪. Q-Q Plot‬‬ ‫ﺑﻌﺾ اﻟﻤﻌﺎﻳﻴﺮ اﻻﺧﺮى ﻹﺧﺘﻴﺎر ﻧﻤﻮذج اﻟﻤﻨﺎﺳﺐ‪:‬‬ ‫‪ (١‬إﺣﺼﺎﺋﻴﺔ آﻴﻮ ﻟـ ﻟﺠﻨﻖ‪-‬ﺑﻮآﺲ ‪ Ljung-Box Q statistc‬وﺗﺨﺘﺼﺮ ‪ LBQ‬وﺗﺴﺘﺨﺪم‬ ‫ﻹﺧﺘﺒﺎر اﻟﻔﺮﺿﻴﺔ‪:‬‬ ‫‪H 0 : ρ1 = ρ 2 = L = ρ K = 0‬‬ ‫وﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‪:‬‬

‫‪٧٤‬‬

‫)‪χ 2 ( K − m‬‬

‫‪rk2‬‬ ‫‪k =1 n − k‬‬ ‫‪K‬‬

‫∑ )‪Q = n (n + 2‬‬

‫ﺣﻴﺚ ‪ m‬ﻋﺪد اﻟﻤﻌﺎﻟﻢ اﻟﻤﻘﺪرة ﻓﻲ اﻟﻨﻤﻮذج‪.‬‬ ‫‪ (٢‬ﻣﻌﻴﺎر اﻹﻋﻼم اﻟﺬاﺗﻲ ‪ Automatic Information Criteria‬وﺗﺨﺘﺼﺮ ‪ AIC‬وﺗﻌﻄﻰ‬ ‫ﺑﺎﻟﻌﻼﻗﺔ‪:‬‬ ‫‪2‬‬ ‫‪AIC ( m ) = n ln σ a + 2m‬‬ ‫ﺣﻴﺚ ‪ m‬ﻋﺪد اﻟﻤﻌﺎﻟﻢ اﻟﻤﻘﺪرة ﻓﻲ اﻟﻨﻤﻮذج وﻧﺨﺘﺎر اﻟﻨﻤﻮذج اﻟﺬي ﻳﻌﻄﻲ‬ ‫) ‪min AIC ( m‬‬ ‫‪m‬‬

‫أﻣﺜﻠﺔ وﺣﺎﻻت دراﺳﺔ ‪: Examples and Case Studies‬‬ ‫‪ -١‬اﻟﺒﻴﺎﻧﺎت اﻟﺘﺎﻟﻴﺔ ﻟﻤﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺸﺎهﺪة‬ ‫)‪z(t‬‬

‫‪59.9315‬‬ ‫‪61.0344‬‬ ‫‪60.6643‬‬ ‫‪60.5777‬‬ ‫‪59.9086‬‬ ‫‪62.2658‬‬ ‫‪61.9013‬‬ ‫‪54.5045‬‬ ‫‪55.8451‬‬ ‫‪59.3572‬‬ ‫‪61.8956‬‬ ‫‪59.7252‬‬ ‫‪58.4651‬‬ ‫‪61.2797‬‬ ‫‪58.0143‬‬ ‫‪56.6085‬‬ ‫‪56.8697‬‬ ‫‪57.2961‬‬ ‫‪59.6266‬‬ ‫‪58.1105‬‬ ‫‪61.9152‬‬ ‫‪59.0778‬‬ ‫‪58.2759‬‬ ‫‪59.5015‬‬ ‫‪63.4124‬‬

‫‪60.7196‬‬ ‫‪62.8556‬‬ ‫‪59.3191‬‬ ‫‪59.5353‬‬ ‫‪59.6004‬‬ ‫‪59.1455‬‬ ‫‪60.9492‬‬ ‫‪58.1690‬‬ ‫‪56.7437‬‬ ‫‪63.8189‬‬ ‫‪59.3402‬‬ ‫‪59.9246‬‬ ‫‪57.9624‬‬ ‫‪57.8803‬‬ ‫‪54.2185‬‬ ‫‪57.0642‬‬ ‫‪57.3940‬‬ ‫‪59.3236‬‬ ‫‪61.1107‬‬ ‫‪58.7336‬‬ ‫‪62.1957‬‬ ‫‪56.9972‬‬ ‫‪61.8685‬‬ ‫‪56.6666‬‬ ‫‪60.7356‬‬

‫‪57.2318‬‬ ‫‪64.6886‬‬ ‫‪60.5820‬‬ ‫‪58.3755‬‬ ‫‪60.1325‬‬ ‫‪58.0151‬‬ ‫‪59.5333‬‬ ‫‪55.7339‬‬ ‫‪58.9585‬‬ ‫‪61.1520‬‬ ‫‪59.0087‬‬ ‫‪60.5289‬‬ ‫‪59.1567‬‬ ‫‪60.3373‬‬ ‫‪55.4219‬‬ ‫‪62.3728‬‬ ‫‪60.1458‬‬ ‫‪57.5307‬‬ ‫‪61.5614‬‬ ‫‪60.0377‬‬ ‫‪60.8256‬‬ ‫‪59.0780‬‬ ‫‪63.1777‬‬ ‫‪56.0309‬‬ ‫‪59.2298‬‬

‫‪56.1346‬‬ ‫‪63.5049‬‬ ‫‪62.4654‬‬ ‫‪59.3054‬‬ ‫‪58.4174‬‬ ‫‪59.0903‬‬ ‫‪58.8802‬‬ ‫‪63.7261‬‬ ‫‪61.8370‬‬ ‫‪61.6337‬‬ ‫‪58.7564‬‬ ‫‪60.8942‬‬ ‫‪56.5413‬‬ ‫‪61.4310‬‬ ‫‪59.2086‬‬ ‫‪60.8605‬‬ ‫‪61.4451‬‬ ‫‪53.8560‬‬ ‫‪59.4119‬‬ ‫‪59.3488‬‬ ‫‪59.3839‬‬ ‫‪59.8597‬‬ ‫‪58.3583‬‬ ‫‪56.1494‬‬ ‫‪61.7218‬‬

‫‪56.4828‬‬ ‫‪63.9622‬‬ ‫‪59.1721‬‬ ‫‪60.9225‬‬ ‫‪58.1483‬‬ ‫‪59.4554‬‬ ‫‪61.7122‬‬ ‫‪61.6708‬‬ ‫‪58.5870‬‬ ‫‪60.2990‬‬ ‫‪58.2273‬‬ ‫‪63.6776‬‬ ‫‪54.6083‬‬ ‫‪62.3827‬‬ ‫‪57.8763‬‬ ‫‪60.3843‬‬ ‫‪63.5907‬‬ ‫‪58.1711‬‬ ‫‪59.9346‬‬ ‫‪58.0423‬‬ ‫‪55.4010‬‬ ‫‪59.0997‬‬ ‫‪59.5097‬‬ ‫‪59.2927‬‬ ‫‪61.1168‬‬

‫‪58.9275‬‬ ‫‪63.1547‬‬ ‫‪57.9813‬‬ ‫‪61.1856‬‬ ‫‪61.8108‬‬ ‫‪59.1609‬‬ ‫‪65.1325‬‬ ‫‪62.6899‬‬ ‫‪57.9363‬‬ ‫‪59.9443‬‬ ‫‪54.7163‬‬ ‫‪60.0538‬‬ ‫‪54.5550‬‬ ‫‪63.3933‬‬ ‫‪59.9569‬‬ ‫‪60.0855‬‬ ‫‪60.6919‬‬ ‫‪61.1852‬‬ ‫‪60.6201‬‬ ‫‪60.7227‬‬ ‫‪58.6501‬‬ ‫‪59.0970‬‬ ‫‪60.3563‬‬ ‫‪60.2513‬‬ ‫‪61.2179‬‬

‫‪59.5257‬‬ ‫‪61.4230‬‬ ‫‪58.7108‬‬ ‫‪61.4761‬‬ ‫‪62.1789‬‬ ‫‪61.7008‬‬ ‫‪60.5918‬‬ ‫‪59.4444‬‬ ‫‪57.3334‬‬ ‫‪62.1017‬‬ ‫‪57.3292‬‬ ‫‪63.1070‬‬ ‫‪58.1895‬‬ ‫‪61.9205‬‬ ‫‪56.2599‬‬ ‫‪62.1362‬‬ ‫‪58.4221‬‬ ‫‪60.8962‬‬ ‫‪60.3030‬‬ ‫‪60.7021‬‬ ‫‪58.5790‬‬ ‫‪57.1459‬‬ ‫‪60.9815‬‬ ‫‪60.2052‬‬ ‫‪60.9013‬‬

‫‪60.1815‬‬ ‫‪61.0640‬‬ ‫‪58.0059‬‬ ‫‪61.2223‬‬ ‫‪61.9753‬‬ ‫‪60.4833‬‬ ‫‪63.4411‬‬ ‫‪59.3478‬‬ ‫‪56.7241‬‬ ‫‪58.1281‬‬ ‫‪61.7840‬‬ ‫‪60.9021‬‬ ‫‪60.7001‬‬ ‫‪61.9462‬‬ ‫‪61.9448‬‬ ‫‪60.9805‬‬ ‫‪57.5151‬‬ ‫‪59.2145‬‬ ‫‪58.5278‬‬ ‫‪60.3550‬‬ ‫‪59.4242‬‬ ‫‪60.3319‬‬ ‫‪61.5555‬‬ ‫‪59.7755‬‬ ‫‪59.4755‬‬

‫اوﻻ ﻧﺮﺳﻢ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻓﻲ ﻣﺨﻄﻂ زﻣﻨﻲ ‪ Time Plot‬ﺑﺈﺳﺘﺨﺪام اﻟﺤﺰﻣﺔ اﻹﺣﺼﺎﺋﻴﺔ ‪MINITAB‬‬ ‫آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫;')‪MTB > TSPlot 'z(t‬‬ ‫>‪SUBC‬‬ ‫;‪Index‬‬ ‫>‪SUBC‬‬ ‫;‪TDisplay 11‬‬ ‫>‪SUBC‬‬ ‫;‪Symbol‬‬ ‫>‪SUBC‬‬ ‫;‪Connect‬‬ ‫>‪SUBC‬‬ ‫‪Title "An obseved Time Series".‬‬

‫‪٧٥‬‬

A n o b s e v e d T im e S e r ie s

z(t)

65

60

55

50

In d e x

100

150

200

‫ﺛﺎﻧﻴﺎ ﻧﺤﺴﺐ وﻧﺮﺳﻢ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﻌﻴﻨﻲ ﺑﺈﺳﺘﺨﺪام اﻷﻣﺮ‬ MTB > %ACF 'z(t)'; SUBC> MAXLAG 20; SUBC> TITLE"SACF of observed Time Series". Executing from file: H:\MTBWIN\MACROS\ACF.MAC

Autocorrelation

S A C F o f o b s e rv e d T im e S e rie s 1 .0 0 .8 0 .6 0 .4 0 .2 0 .0 -0 .2 -0 .4 -0 .6 -0 .8 -1 .0

5

L a g C o rr

T

LBQ

0 .5 1 0 .2 0 -0 .0 0 -0 .0 5 -0 .0 8 -0 .1 8 -0 .1 9

7 .1 9 2 .3 2 - 0 .0 1 - 0 .5 9 - 0 .9 5 - 2 .0 5 - 2 .0 9

5 2 .4 8 6 0 .7 8 6 0 .7 8 6 1 .3 4 6 2 .8 2 6 9 .9 2 7 7 .5 8

1 2 3 4 5 6 7

10

15

L a g C o rr

T

LBQ

- 0 .1 4 - 0 .1 4 - 0 .0 9 - 0 .0 7 - 0 .0 8 - 0 .0 2 0 .0 3

-1 .5 0 -1 .5 2 -0 .9 0 -0 .7 1 -0 .7 9 -0 .2 1 0 .3 2

8 1 .7 6 8 6 .1 4 8 7 .7 3 8 8 .7 3 8 9 .9 7 9 0 .0 5 9 0 .2 7

8 9 10 11 12 13 14

20

L a g C o rr

T

LBQ

0 .0 7 0 .1 3 0 .1 7 0 .2 0 0 .1 2 0 .0 6

0 .6 8 1 .3 3 1 .7 5 2 .0 6 1 .2 1 0 .6 1

9 1 .2 3 9 4 .8 6 1 0 1 .3 3 1 1 0 .6 3 1 1 3 .9 8 1 1 4 .8 6

15 16 17 18 19 20

‫ﺛﺎﻟﺜﺎ ﻧﺤﺴﺐ وﻧﺮﺳﻢ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻲ ﺑﺈﺳﺘﺨﺪام اﻷﻣﺮ‬ MTB > %PACF 'z(t)'; SUBC> MAXLAG 20; SUBC> TITLE"SPACF of obseved Time Series". Executing from file: H:\MTBWIN\MACROS\PACF.MAC

٧٦

Partial Autocorrelation

S P A C F o f o b se ve d T im e S e rie s 1 .0 0 .8 0 .6 0 .4 0 .2 0 .0 -0 .2 -0 .4 -0 .6 -0 .8 -1 .0

5

10

L a g P AC

T

0 .5 1 -0 .0 8 -0 .1 0 0 .0 0 -0 .0 5 -0 .1 6 -0 .0 4

7 .1 9 -1 .0 7 -1 .3 8 0 .0 4 -0 .7 3 -2 .3 4 -0 .5 0

1 2 3 4 5 6 7

15

L a g P AC

T

-0 .0 1 -0 .1 2 0 .0 1 -0 .0 4 -0 .0 9 0 .0 3 0 .0 2

-0 .1 2 -1 .6 3 0 .1 6 -0 .6 0 -1 .3 4 0 .3 9 0 .3 2

8 9 10 11 12 13 14

20

L a g P AC

T

-0 .0 2 0 .0 9 0 .0 8 0 .0 6 -0 .0 3 0 .0 4

-0 .2 3 1 .2 8 1 .2 0 0 .8 6 -0 .4 5 0 .5 2

15 16 17 18 19 20

‫ﻣﻦ أﻧﻤﺎط اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ و اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻲ ﻧﻼﺣﻆ ان اﻟﻤﺘﺴﻠﺴﻠﺔ ﺗﺘﺒﻊ ﻧﻤﻮذج‬ ‫ وﻟﻬﺬا ﻧﻄﺒﻖ اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح ﻋﻠﻲ اﻟﻤﺸﺎهﺪات ﺑﺈﺳﺘﺨﺪام اﻷﻣﺮ‬AR(١) MTB > Name c7 = 'RESI1' MTB > ARIMA 1 0 0 'z(t)' 'RESI1'; SUBC> Constant; SUBC> Forecast 5 c4 c5 c6; SUBC> GACF; SUBC> GPACF; SUBC> GHistogram; SUBC> GNormalplot.

ARIMA Model ARIMA model for z(t) Estimates at each iteration Iteration SSE Parameters 0 839.667 0.100 53.870 1 746.819 0.250 44.876 2 695.840 0.400 35.883 3 685.086 0.502 29.769 4 685.054 0.507 29.458 5 685.054 0.507 29.443 Relative change in each estimate less than Final Estimates of Parameters Type Coef StDev AR 1 0.5073 0.0611 Constant 29.4429 0.1309 Mean 59.7571 0.2656

0.0010

T 8.30 224.98

Number of observations: 201 Residuals: SS = 685.020 (backforecasts excluded) MS = 3.442 DF = 199 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 10.8(DF=11) 27.6(DF=23) 35.9(DF=35)

48 45.0(DF=47)

Forecasts from period 201 Period 202 203

Forecast 59.7079 59.7322

95 Percent Limits Lower Upper 56.0707 63.3451 55.6537 63.8106

٧٧

Actual

204 205 206

59.7445 59.7507 59.7539

55.5600 55.5394 55.5357

zt = 59.76 + 0.51( zt −1 − 59.76) + at , at

63.9290 63.9620 63.9721

:‫وﻧﺴﺘﻨﺘﺞ اﻟﺘﺎﻟﻲ‬ ‫ اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح هﻮ‬-١

WN ( 0,3.44 )

:‫ ﻟﻬﺎ هﻲ‬t ‫ اﻟﻤﻌﺎﻟﻢ اﻟﻤﻘﺪرة وإﻧﺤﺮاﻓﻬﺎ اﻟﻤﻌﻴﺎري و ﻗﻴﻤﺔ‬-٢

( )

φˆ1 = 0.51, s.e. φˆ1 = 0.061, t = 8.3 µˆ = 59.76, s.e. ( µˆ ) = 0.66

( )

δˆ = 29.44, s.e. δˆ = 0.131, t = 224.98 σˆ 2 = 3.44, with d . f . = 199

:‫راﺑﻌﺎ ﻧﻔﺤﺺ اﻟﺒﻮاﻗﻲ‬ ‫ إﺧﺘﺒﺎر ﻣﺘﻮﺳﻂ اﻟﺒﻮاﻗﻲ‬-١ MTB > ZTest 0.0 1.855 'RESI1'; SUBC> Alternative 0; SUBC> GHistogram; SUBC> GDotplot; SUBC> GBoxplot.

Z-Test Test of mu = 0.000 vs mu not = 0.000 The assumed sigma = 1.85 Variable RESI1

N 201

Mean -0.002

StDev 1.851

SE Mean 0.131

Z -0.01

P 0.99

‫ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ اﻟﺼﻔﺮﻳﺔ ﺑﺄن اﻟﻤﺘﻮﺳﻂ ﻳﺴﺎوي اﻟﺼﻔﺮ‬ ‫ إﺧﺘﺒﺎر ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬-٢ MTB > Runs 0 'RESI1'.

Runs Test RESI1 K =

0.0000

The observed number of runs = 94 The expected number of runs = 101.0796 107 Observations above K 94 below The test is significant at 0.3149 Cannot reject at alpha = 0.05

‫ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ اﻟﺼﻔﺮﻳﺔ ﺑﺄن اﻟﺒﻮاﻗﻲ ﻋﺸﻮاﺋﻴﺔ‬ ‫ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﺒﻮاﻗﻲ‬-٣

٧٨

ACF of Residuals for z(t) (with 95% confidence limits for the autocorrelations) 1.0 0.8

Autocorrelation

0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0 5

10

15

20

25

30

35

40

45

50

Lag

‫ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﺒﻮاﻗﻲ‬-٤ PACF of Residuals for z(t) (with 95% confidence limits for the partial autocorrelations) 1.0

Partial Autocorrelation

0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0 5

10

15

20

25

30

35

40

45

50

Lag

‫ﻧﻼﺣﻆ ان أﻧﻤﺎط اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﺗﺘﺒﻊ أﻧﻤﺎط ﻣﺘﺴﻠﺴﻠﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‬ : ‫ إﺧﺘﺒﺎر ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬-٥ ‫ ﻧﺮﺳﻢ اﻟﻤﻀﻠﻊ اﻟﺘﻜﺮاري ﻟﻠﺒﻮاﻗﻲ‬-‫ا‬ Histogram of the Residuals (response is z(t))

Frequency

30

20

10

0 -5

0

Residual

٧٩

5

:‫ وهﺬا ﻻﻳﻜﻔﻲ ﺑﻞ ﻳﺠﺐ ان ﻧﻨﻈﺮ اﻟﻰ‬.‫ﻧﻼﺣﻆ أﻧﻪ ﻣﺘﻨﺎﻇﺮ وﻟﺔ ﺷﻜﻞ اﻟﺘﻮزﻳﻊ اﻟﻄﺒﻴﻌﻲ ﺗﻘﺮﻳﺒﺎ‬ Normal Probability Plot ‫ رﺳﻢ اﻻﺣﺘﻤﺎل اﻟﻄﺒﻴﻌﻲ‬-‫ب‬ Normal Probability Plot for RESI1

99

Mean:

-1.6E-03

StDev:

1.85070

95 90

Percent

80 70 60 50 40 30 20 10 5

1

-5.0

-2.5

0.0

2.5

5.0

Data

:‫واﺿﺢ ﻣﻦ اﻟﺮﺳﻢ أن اﻟﺒﻮاﻗﻲ ﻃﺒﻴﻌﻴﺔ وﻟﻠﺘﺄآﺪ ﻧﻘﻮم‬ ‫ ﻟﻄﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬K-S Test ‫ ﺑﺈﺧﺘﺒﺎر‬-‫ج‬ MTB > %NormPlot 'RESI1'; SUBC> Kstest; SUBC> Title "Normal Test for Residuals". Executing from file: H:\MTBWIN\MACROS\NormPlot.MAC

Normal Test for Residuals

.999 .99

Probability

.95 .80 .50 .20 .05 .01 .001 -5

0

5

RESI1 Average: -0.0016272 StDev: 1.85070 N: 201

H 0 : Residuals

Kolmogorov-Smirnov Normality Test D+: 0.045 D-: 0.060 D : 0.060 Approximate P-Value: 0.074

:‫وﻧﻼﺣﻆ اﻟﺘﺎﻟﻲ‬ ‫اﻹﺧﺘﺒﺎر هﻮ‬

N ( 0,3.44 )

H1 : Residuals§ N ( 0,3.44 )

‫ﺳﻤﻴﺮﻧﻮف اﻋﻄﻰ‬-‫إﺧﺘﺒﺎر آﻮﻟﻤﻮﺟﻮروف‬ +



D = 0.045, D = 0.06, D = 0.06 ‫ أي اﻧﻨﺎ ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ‬α = 0.05 ‫ وهﻲ أآﺒﺮ ﻣﻦ‬٠٫٠٧٤ ‫ ﻟﻺﺧﺘﺒﺎر هﻲ‬P-Value ‫اﻟـ‬

.‫اﻟﺼﻔﺮﻳﺔ‬ :‫ﺗﻮﻟﻴﺪ ﺗﻨﺒﺆات‬ ٨٠

‫ ﻗﻴﻢ ﻣﺴﺘﻘﺒﻠﻴﺔ‬٥ ‫اﺳﺘﺨﺪﻣﻨﺎ اﻟﻨﻤﻮذج ﻟﻠﺘﻨﺒﺆ ﻋﻦ‬ Forecasts from period 201 Period 202 203 204 205 206

Forecast 59.7079 59.7322 59.7445 59.7507 59.7539

95 Percent Limits Lower Upper 56.0707 63.3451 55.6537 63.8106 55.5600 63.9290 55.5394 63.9620 55.5357 63.9721

Actual

‫وﻧﺮﺳﻤﻬﺎ ﺑﺎﻷﻣﺮ اﻟﺘﺎﻟﻲ‬ Plot C4*C8 C5*C8 C6*C8; SUBC> Connect; SUBC> Type 1; SUBC> Color 1; SUBC> Size 1; SUBC> Title "Forecast of 5 future value with 95% limits"; SUBC> Overlay.

Forecast of 5 future value with 95% limits 64 63 62

C4

61 60 59 58 57 56 55 1

2

3

4

5

C8

.‫واﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻳﻌﻄﻲ اﻟﺠﺰء اﻷﺧﻴﺮ ﻣﻦ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻣﻊ اﻟﺘﻨﺒﺆات وﻓﺘﺮات اﻟﺘﻨﺒﺆ‬ Forecast of 5 future value with 95% limits 64 63 62

C9

61 60 59 58 57 56 55 180

190

200

C8

‫ اﻟﺒﻴﺎﻧﺎت اﻟﺘﺎﻟﻴﺔ ﻟﻤﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺸﺎهﺪة‬-٢ ٨١

‫‪494.948‬‬ ‫‪503.170‬‬ ‫‪496.806‬‬ ‫‪504.340‬‬ ‫‪496.372‬‬ ‫‪503.506‬‬ ‫‪496.062‬‬ ‫‪501.521‬‬ ‫‪507.886‬‬ ‫‪506.345‬‬ ‫‪507.735‬‬ ‫‪497.295‬‬ ‫‪497.642‬‬ ‫‪510.287‬‬ ‫‪496.029‬‬ ‫‪503.134‬‬ ‫‪501.634‬‬ ‫‪499.611‬‬ ‫‪500.388‬‬ ‫‪505.987‬‬ ‫‪493.858‬‬ ‫‪494.857‬‬ ‫‪497.861‬‬ ‫‪493.017‬‬ ‫‪498.931‬‬ ‫‪497.794‬‬ ‫‪513.900‬‬ ‫‪504.717‬‬ ‫‪502.209‬‬ ‫‪497.992‬‬ ‫‪501.387‬‬

‫‪496.208‬‬ ‫‪498.429‬‬ ‫‪496.227‬‬ ‫‪491.923‬‬ ‫‪499.260‬‬ ‫‪500.356‬‬ ‫‪494.033‬‬ ‫‪497.393‬‬ ‫‪495.431‬‬ ‫‪496.285‬‬ ‫‪507.673‬‬ ‫‪503.415‬‬ ‫‪499.783‬‬ ‫‪496.181‬‬ ‫‪497.133‬‬ ‫‪494.878‬‬ ‫‪497.943‬‬ ‫‪499.299‬‬ ‫‪498.787‬‬ ‫‪499.467‬‬ ‫‪502.933‬‬ ‫‪495.430‬‬ ‫‪511.741‬‬ ‫‪501.999‬‬ ‫‪496.078‬‬ ‫‪497.413‬‬ ‫‪495.353‬‬ ‫‪499.134‬‬ ‫‪503.130‬‬ ‫‪504.174‬‬ ‫‪501.701‬‬

‫‪507.382‬‬ ‫‪502.233‬‬ ‫‪505.884‬‬ ‫‪496.665‬‬ ‫‪500.074‬‬ ‫‪492.286‬‬ ‫‪508.489‬‬ ‫‪504.965‬‬ ‫‪500.664‬‬ ‫‪507.072‬‬ ‫‪485.991‬‬ ‫‪500.921‬‬ ‫‪503.852‬‬ ‫‪498.380‬‬ ‫‪498.060‬‬ ‫‪500.199‬‬ ‫‪504.785‬‬ ‫‪505.517‬‬ ‫‪505.475‬‬ ‫‪497.462‬‬ ‫‪504.817‬‬ ‫‪504.336‬‬ ‫‪490.887‬‬ ‫‪503.465‬‬ ‫‪506.649‬‬ ‫‪505.218‬‬ ‫‪498.477‬‬ ‫‪496.680‬‬ ‫‪491.202‬‬ ‫‪495.699‬‬ ‫‪492.716‬‬

‫‪498.440‬‬ ‫‪496.678‬‬ ‫‪493.371‬‬ ‫‪506.329‬‬ ‫‪498.598‬‬ ‫‪516.373‬‬ ‫‪499.217‬‬ ‫‪495.000‬‬ ‫‪492.352‬‬ ‫‪495.423‬‬ ‫‪505.577‬‬ ‫‪501.819‬‬ ‫‪501.175‬‬ ‫‪504.666‬‬ ‫‪509.407‬‬ ‫‪504.408‬‬ ‫‪501.417‬‬ ‫‪492.318‬‬ ‫‪496.757‬‬ ‫‪498.403‬‬ ‫‪491.707‬‬ ‫‪505.429‬‬ ‫‪514.220‬‬ ‫‪502.414‬‬ ‫‪491.418‬‬ ‫‪496.150‬‬ ‫‪498.016‬‬ ‫‪501.542‬‬ ‫‪507.590‬‬ ‫‪497.647‬‬ ‫‪504.640‬‬

‫‪488.539‬‬ ‫‪506.040‬‬ ‫‪501.605‬‬ ‫‪497.785‬‬ ‫‪502.891‬‬ ‫‪491.981‬‬ ‫‪493.161‬‬ ‫‪505.581‬‬ ‫‪501.064‬‬ ‫‪497.883‬‬ ‫‪500.744‬‬ ‫‪493.866‬‬ ‫‪495.868‬‬ ‫‪494.885‬‬ ‫‪494.814‬‬ ‫‪495.954‬‬ ‫‪493.552‬‬ ‫‪497.130‬‬ ‫‪497.626‬‬ ‫‪506.259‬‬ ‫‪504.085‬‬ ‫‪490.683‬‬ ‫‪487.344‬‬ ‫‪494.614‬‬ ‫‪507.438‬‬ ‫‪512.122‬‬ ‫‪505.367‬‬ ‫‪504.012‬‬ ‫‪497.231‬‬ ‫‪505.234‬‬ ‫‪496.877‬‬

‫‪511.026‬‬ ‫‪489.348‬‬ ‫‪498.229‬‬ ‫‪502.545‬‬ ‫‪503.107‬‬ ‫‪496.830‬‬ ‫‪507.020‬‬ ‫‪495.355‬‬ ‫‪504.712‬‬ ‫‪504.072‬‬ ‫‪495.737‬‬ ‫‪503.580‬‬ ‫‪501.700‬‬ ‫‪507.582‬‬ ‫‪501.928‬‬ ‫‪503.325‬‬ ‫‪500.484‬‬ ‫‪502.066‬‬ ‫‪501.082‬‬ ‫‪493.843‬‬ ‫‪498.571‬‬ ‫‪501.703‬‬ ‫‪498.599‬‬ ‫‪500.256‬‬ ‫‪497.643‬‬ ‫‪490.619‬‬ ‫‪500.324‬‬ ‫‪499.779‬‬ ‫‪497.956‬‬ ‫‪502.681‬‬ ‫‪502.823‬‬

‫‪496.650‬‬ ‫‪501.649‬‬ ‫‪502.969‬‬ ‫‪506.459‬‬ ‫‪494.559‬‬ ‫‪500.508‬‬ ‫‪498.319‬‬ ‫‪498.304‬‬ ‫‪501.527‬‬ ‫‪494.833‬‬ ‫‪504.129‬‬ ‫‪507.594‬‬ ‫‪498.294‬‬ ‫‪496.932‬‬ ‫‪504.394‬‬ ‫‪499.982‬‬ ‫‪498.640‬‬ ‫‪502.173‬‬ ‫‪504.346‬‬ ‫‪504.913‬‬ ‫‪500.712‬‬ ‫‪504.204‬‬ ‫‪506.403‬‬ ‫‪502.721‬‬ ‫‪503.746‬‬ ‫‪506.001‬‬ ‫‪501.827‬‬ ‫‪496.063‬‬ ‫‪505.709‬‬ ‫‪493.502‬‬ ‫‪497.421‬‬ ‫‪497.048‬‬

‫)‪z(t‬‬ ‫‪499.148‬‬ ‫‪503.975‬‬ ‫‪498.758‬‬ ‫‪493.057‬‬ ‫‪499.890‬‬ ‫‪507.416‬‬ ‫‪498.090‬‬ ‫‪504.877‬‬ ‫‪494.918‬‬ ‫‪499.173‬‬ ‫‪496.765‬‬ ‫‪482.567‬‬ ‫‪500.989‬‬ ‫‪501.331‬‬ ‫‪489.314‬‬ ‫‪502.720‬‬ ‫‪502.489‬‬ ‫‪495.691‬‬ ‫‪496.252‬‬ ‫‪499.279‬‬ ‫‪498.169‬‬ ‫‪497.015‬‬ ‫‪503.195‬‬ ‫‪500.252‬‬ ‫‪498.158‬‬ ‫‪500.409‬‬ ‫‪496.225‬‬ ‫‪491.726‬‬ ‫‪489.032‬‬ ‫‪500.024‬‬ ‫‪505.194‬‬ ‫‪499.574‬‬

‫ﺍﻭﻻ ‪ :‬ﺴﻭﻑ ﻨﺭﺴﻡ ﻓﻘﻁ ‪ ٥٠‬ﻤﺸﺎﻫﺩﺓ ﻤﻥ ﻫﺫﻩ ﺍﻟﻤﺘﺴﻠﺴﻠﺔ‬

‫‪510‬‬

‫)‪z(t‬‬

‫‪500‬‬

‫‪490‬‬

‫‪50‬‬

‫‪40‬‬

‫‪30‬‬

‫ﺛﺎﻧﻴﺎ‪ :‬ﻧﺤﺴﺐ وﻧﺮﺳﻢ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﻌﻴﻨﻲ‪:‬‬

‫‪٨٢‬‬

‫‪20‬‬

‫‪10‬‬

‫‪In d e x‬‬

Autocorrelation

Autocorrelation Function for z(t) 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

5

Lag Corr

10

T

LBQ

1 -0.53 -8.38 2 -0.05 -0.64 3 0.12 1.52 4 0.04 0.52 5 -0.13 -1.61 6 0.11 1.39 7 -0.09 -1.13

71.03 71.68 75.39 75.83 80.08 83.33 85.52

Lag Corr

T

15

LBQ

8 0.10 1.19 87.98 9 -0.11 -1.31 91.01 10 0.15 1.77 96.64 11 -0.18 -2.16 105.32 12 0.11 1.25 108.35 13 0.05 0.56 108.97 14 -0.12 -1.36 112.61

Lag Corr

20

T

LBQ

15 0.05 0.60 113.33 16 -0.02 -0.26 113.47 17 0.07 0.82 114.85 18 -0.10 -1.16 117.65 19 0.07 0.75 118.81 20 -0.06 -0.73 119.93

:‫ ﻧﺤﺴﺐ وﻧﺮﺳﻢ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻲ‬:‫ﺛﺎﻟﺜﺎ‬ Partial Autocorrelation

Partial Autocorrelation Function for z(t) 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

5

10

15

Lag PAC

T

Lag PAC

T

1 -0.53 2 -0.46 3 -0.29 4 -0.07 5 -0.11 6 0.02 7 -0.09

-8.38 -7.28 -4.52 -1.08 -1.71 0.30 -1.36

8 0.04 9 -0.07 10 0.12 11 -0.07 12 -0.04 13 0.08 14 -0.04

0.65 -1.09 1.94 -1.15 -0.71 1.19 -0.59

Lag PAC

20

T

15 0.03 0.49 16 -0.13 -2.09 17 0.06 1.01 18 -0.08 -1.32 19 0.03 0.44 20 -0.13 -2.07

‫ وﺑﺘﻄﺒﻴﻖ هﺬا اﻟﻨﻤﻮذج ﻧﺠﺪ‬MA(١) ‫ﻣﻦ اﻷﻧﻤﺎط اﻟﻤﺸﺎهﺪة ﻧﻼﺣﻆ ان اﻟﻤﺘﺴﻠﺴﺔ ﻗﺪ ﺗﺘﺒﻊ ﻧﻤﻮذج‬ MTB > Name c7 = 'RESI1' MTB > ARIMA 0 0 1 'z(t)' 'RESI1'; SUBC> Constant; SUBC> Forecast 5 c4 c5 c6; SUBC> GACF; SUBC> GPACF; SUBC> GHistogram; SUBC> GNormalplot.

ARIMA Model ARIMA model for z(t) Estimates at each iteration Iteration SSE Parameters 0 6081.19 0.100 500.046 1 5265.34 0.250 500.004 2 4615.22 0.400 499.980 3 4109.70 0.550 499.967 4 3766.60 0.700 499.960 5 3727.32 0.841 499.959 6 3687.70 0.797 499.963

٨٣

7 3687.08 0.790 499.962 8 3687.07 0.791 499.962 9 3687.07 0.790 499.962 Relative change in each estimate less than Final Estimates of Parameters Type Coef StDev MA 1 0.7905 0.0386 Constant 499.962 0.051 Mean 499.962 0.051

0.0010

T 20.50 9708.40

Number of observations: 250 Residuals: SS = 3684.13 (backforecasts excluded) MS = 14.86 DF = 248 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 26.7(DF=11) 35.9(DF=23) 63.1(DF=35)

48 82.8(DF=47)

Forecasts from period 250 Period 251 252 253 254 255

Forecast 502.256 499.962 499.962 499.962 499.962

95 Percent Limits Lower Upper 494.700 509.812 490.330 509.593 490.330 509.593 490.330 509.593 490.330 509.593

zt = 499.962 + at − 0.7905at −1 , at

WN ( 0,14.86 )

Actual

:‫وﻧﺴﺘﻨﺘﺞ اﻟﺘﺎﻟﻲ‬ ‫ اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح هﻮ‬-١

:‫ ﻟﻬﺎ هﻲ‬t ‫ اﻟﻤﻌﺎﻟﻢ اﻟﻤﻘﺪرة وإﻧﺤﺮاﻓﻬﺎ اﻟﻤﻌﻴﺎري و ﻗﻴﻤﺔ‬-٢

( )

θˆ1 = 0.7905, s.e. θˆ1 = 0.0386, t = 20.50

( )

µˆ = δˆ = 499.962, s.e. δˆ = 0.051, t = 9708.40 σˆ 2 = 14.86, with d . f . = 248

:‫راﺑﻌﺎ ﻧﻔﺤﺺ اﻟﺒﻮاﻗﻲ‬ ‫ إﺧﺘﺒﺎر ﻣﺘﻮﺳﻂ اﻟﺒﻮاﻗﻲ‬-١ MTB > ZTest 0.0 3.847 'RESI1'; SUBC> Alternative 0.

Z-Test Test of mu = 0.000 vs mu not = 0.000 The assumed sigma = 3.85 Variable RESI1

N 250

Mean -0.007

StDev 3.847

SE Mean 0.243

Z -0.03

P 0.98

‫ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ اﻟﺼﻔﺮﻳﺔ ﺑﺄن اﻟﻤﺘﻮﺳﻂ ﻳﺴﺎوي اﻟﺼﻔﺮ‬ ‫ إﺧﺘﺒﺎر ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬-٢ MTB > Runs 0 'RESI1'.

Runs Test RESI1

٨٤

K =

0.0000

The observed number of runs = 134 The expected number of runs = 125.9920 126 Observations above K 124 below The test is significant at 0.3103 Cannot reject at alpha = 0.05

‫ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ اﻟﺼﻔﺮﻳﺔ ﺑﺄن اﻟﺒﻮاﻗﻲ ﻋﺸﻮاﺋﻴﺔ‬ ‫ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﺒﻮاﻗﻲ‬-٣ A C F o f R e s id u a ls fo r z (t) ( w ith 9 5 % c o n f id e n c e l im it s f o r th e a u to c o r r e l a tio n s ) 1 .0

0 .8

Autocorrelation

0 .6 0 .4

0 .2 0 .0 -0 .2 -0 .4

-0 .6 -0 .8

-1 .0 5

10

15

20

25

30

35

40

45

50

55

60

Lag

‫ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﺒﻮاﻗﻲ‬-٤ P A C F o f R e s id u a ls f o r z (t) ( w it h 9 5 % c o n f id e n c e l im it s f o r t h e p a r t ia l a u t o c o r r e l a t io n s ) 1 .0

Partial Autocorrelation

0 .8 0 .6 0 .4 0 .2 0 .0 -0 .2 -0 .4 -0 .6 -0 .8 -1 .0 5

10

15

20

25

30

35

40

45

50

55

60

Lag

‫ﻧﻼﺣﻆ ان أﻧﻤﺎط اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﺗﺘﺒﻊ أﻧﻤﺎط ﻣﺘﺴﻠﺴﻠﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‬ : ‫ إﺧﺘﺒﺎر ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬-٥ ‫ ﻧﺮﺳﻢ اﻟﻤﻀﻠﻊ اﻟﺘﻜﺮاري ﻟﻠﺒﻮاﻗﻲ‬-‫ا‬

٨٥

H istogram of the R esiduals (resp on se is z(t))

Frequency

30

20

10

0 -10

0

10

R esidual

:‫ وهﺬا ﻻﻳﻜﻔﻲ ﺑﻞ ﻳﺠﺐ ان ﻧﻨﻈﺮ اﻟﻰ‬.‫ﻧﻼﺣﻆ أﻧﻪ ﻣﺘﻨﺎﻇﺮ وﻟﺔ ﺷﻜﻞ اﻟﺘﻮزﻳﻊ اﻟﻄﺒﻴﻌﻲ ﺗﻘﺮﻳﺒﺎ‬ Normal Probability Plot ‫ رﺳﻢ اﻻﺣﺘﻤﺎل اﻟﻄﺒﻴﻌﻲ‬-‫ب‬ Normal Probability Plot for RESI1

99

Mean:

-6.9E-03

StDev:

3.84651

95 90

Percent

80 70 60 50 40 30 20 10 5

1

-10

-5

0

5

10

Data

:‫واﺿﺢ ﻣﻦ اﻟﺮﺳﻢ أن اﻟﺒﻮاﻗﻲ ﻃﺒﻴﻌﻴﺔ وﻟﻠﺘﺄآﺪ ﻧﻘﻮم‬ ‫ ﻟﻄﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬K-S Test ‫ ﺑﺈﺧﺘﺒﺎر‬-‫ج‬ MTB > %Qqplot 'RESI1'; SUBC> Conf 95; SUBC> Ci. Executing from file: H:\MTBWIN\MACROS\Qqplot.MAC

Distribution Function Analysis Normal Dist. Parameter Estimates Data Mean: StDev:

: RESI1 -6.9E-03 3.84651

MTB > %NormPlot 'RESI1'; SUBC> Kstest. Executing from file: H:\MTBWIN\MACROS\NormPlot.MAC

٨٦

Normal Probability Plot

.999 .99

Probability

.95 .80 .50 .20 .05 .01 .001 -10

0

10

RESI1 Average: -0.0069004 StDev: 3.84651 N: 250

Kolmogorov-Smirnov Normality Test D+: 0.034 D-: 0.051 D : 0.051 Approximate P-Value: 0.105

‫ﺳﻤﻴﺮﻧﻮف اﻋﻄﻰ‬-‫إﺧﺘﺒﺎر آﻮﻟﻤﻮﺟﻮروف‬ +



D = 0.034, D = 0.051, D = 0.051 ‫ أي اﻧﻨﺎ ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ ان‬α = 0.05 ‫ وهﻲ أآﺒﺮ ﻣﻦ‬٠٫١٠٥ ‫ ﻟﻺﺧﺘﺒﺎر هﻲ‬P-Value ‫اﻟـ‬

.‫اﻟﺒﻮاﻗﻲ ﻣﻮزﻋﺔ ﻃﺒﻴﻌﻴﺎ‬ :‫ﺗﻮﻟﻴﺪ ﺗﻨﺒﺆات‬ ‫ ﻗﻴﻢ ﻣﺴﺘﻘﺒﻠﻴﺔ‬٥ ‫اﺳﺘﺨﺪﻣﻨﺎ اﻟﻨﻤﻮذج ﻟﻠﺘﻨﺒﺆ ﻋﻦ‬ Forecasts from period 250 Period 251 252 253 254 255

95 Percent Limits Lower Upper 494.700 509.812 490.330 509.593 490.330 509.593 490.330 509.593 490.330 509.593

Forecast 502.256 499.962 499.962 499.962 499.962

Actual

‫ ﺗﻨﺒﺆ‬٩٥% ‫واﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻳﻌﻄﻲ اﻟﺘﻨﺒﺆات ﻣﻊ ﻓﺘﺮات‬ Forecast of 5 future values with 95% limits

C4

510

500

490 1

2

3

4

5

C8

‫ اﻟﺒﻴﺎﻧﺎت اﻟﺘﺎﻟﻴﺔ ﻟﻤﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ﻣﺸﺎهﺪة‬-٣ ٨٧

‫‪224.678‬‬ ‫‪227.500‬‬ ‫‪221.632‬‬ ‫‪234.178‬‬ ‫‪233.005‬‬ ‫‪232.432‬‬ ‫‪228.189‬‬ ‫‪231.718‬‬ ‫‪222.468‬‬ ‫‪231.163‬‬ ‫‪226.479‬‬ ‫‪234.207‬‬ ‫‪230.455‬‬ ‫‪233.640‬‬ ‫‪233.725‬‬ ‫‪233.089‬‬ ‫‪228.655‬‬ ‫‪220.958‬‬ ‫‪231.531‬‬ ‫‪231.363‬‬ ‫‪228.285‬‬ ‫‪229.320‬‬ ‫‪235.587‬‬ ‫‪228.729‬‬ ‫‪232.700‬‬ ‫‪240.884‬‬ ‫‪236.017‬‬ ‫‪229.321‬‬ ‫‪228.141‬‬ ‫‪228.958‬‬ ‫‪226.114‬‬

‫‪226.760‬‬ ‫‪224.063‬‬ ‫‪225.663‬‬ ‫‪238.463‬‬ ‫‪232.209‬‬ ‫‪230.102‬‬ ‫‪231.121‬‬ ‫‪229.848‬‬ ‫‪223.115‬‬ ‫‪228.418‬‬ ‫‪224.483‬‬ ‫‪232.738‬‬ ‫‪228.163‬‬ ‫‪232.232‬‬ ‫‪236.562‬‬ ‫‪233.427‬‬ ‫‪225.484‬‬ ‫‪220.782‬‬ ‫‪232.030‬‬ ‫‪230.938‬‬ ‫‪229.757‬‬ ‫‪230.636‬‬ ‫‪230.713‬‬ ‫‪230.360‬‬ ‫‪234.625‬‬ ‫‪241.821‬‬ ‫‪230.865‬‬ ‫‪226.222‬‬ ‫‪225.985‬‬ ‫‪229.352‬‬ ‫‪228.837‬‬

‫‪226.641‬‬ ‫‪222.515‬‬ ‫‪229.136‬‬ ‫‪239.577‬‬ ‫‪227.621‬‬ ‫‪229.564‬‬ ‫‪231.633‬‬ ‫‪227.982‬‬ ‫‪227.859‬‬ ‫‪228.618‬‬ ‫‪228.989‬‬ ‫‪229.792‬‬ ‫‪225.447‬‬ ‫‪234.825‬‬ ‫‪235.224‬‬ ‫‪233.044‬‬ ‫‪224.207‬‬ ‫‪222.819‬‬ ‫‪232.315‬‬ ‫‪233.127‬‬ ‫‪229.733‬‬ ‫‪227.948‬‬ ‫‪227.046‬‬ ‫‪230.662‬‬ ‫‪234.107‬‬ ‫‪238.762‬‬ ‫‪230.499‬‬ ‫‪225.771‬‬ ‫‪225.358‬‬ ‫‪229.369‬‬ ‫‪232.881‬‬

‫‪226.778‬‬ ‫‪221.562‬‬ ‫‪232.308‬‬ ‫‪232.653‬‬ ‫‪222.156‬‬ ‫‪229.331‬‬ ‫‪231.319‬‬ ‫‪225.734‬‬ ‫‪230.122‬‬ ‫‪228.225‬‬ ‫‪235.122‬‬ ‫‪228.198‬‬ ‫‪224.928‬‬ ‫‪234.707‬‬ ‫‪231.665‬‬ ‫‪229.575‬‬ ‫‪224.747‬‬ ‫‪226.865‬‬ ‫‪232.027‬‬ ‫‪233.852‬‬ ‫‪232.160‬‬ ‫‪224.258‬‬ ‫‪228.606‬‬ ‫‪234.243‬‬ ‫‪231.265‬‬ ‫‪233.628‬‬ ‫‪226.994‬‬ ‫‪225.616‬‬ ‫‪225.291‬‬ ‫‪230.296‬‬ ‫‪236.029‬‬

‫‪229.903‬‬ ‫‪222.482‬‬ ‫‪236.488‬‬ ‫‪223.408‬‬ ‫‪218.067‬‬ ‫‪229.359‬‬ ‫‪234.668‬‬ ‫‪225.721‬‬ ‫‪230.888‬‬ ‫‪229.851‬‬ ‫‪235.024‬‬ ‫‪226.724‬‬ ‫‪227.812‬‬ ‫‪233.841‬‬ ‫‪228.922‬‬ ‫‪227.089‬‬ ‫‪228.954‬‬ ‫‪232.152‬‬ ‫‪232.135‬‬ ‫‪232.684‬‬ ‫‪235.038‬‬ ‫‪223.994‬‬ ‫‪228.203‬‬ ‫‪239.883‬‬ ‫‪229.486‬‬ ‫‪230.342‬‬ ‫‪223.795‬‬ ‫‪226.271‬‬ ‫‪223.250‬‬ ‫‪228.920‬‬ ‫‪235.339‬‬

‫‪230.260‬‬ ‫‪223.390‬‬ ‫‪236.033‬‬ ‫‪217.433‬‬ ‫‪217.123‬‬ ‫‪235.744‬‬ ‫‪235.767‬‬ ‫‪224.927‬‬ ‫‪228.472‬‬ ‫‪228.227‬‬ ‫‪236.659‬‬ ‫‪225.196‬‬ ‫‪231.266‬‬ ‫‪233.891‬‬ ‫‪230.327‬‬ ‫‪227.032‬‬ ‫‪232.820‬‬ ‫‪232.461‬‬ ‫‪228.727‬‬ ‫‪231.411‬‬ ‫‪233.799‬‬ ‫‪225.273‬‬ ‫‪231.898‬‬ ‫‪243.963‬‬ ‫‪225.070‬‬ ‫‪227.643‬‬ ‫‪223.381‬‬ ‫‪230.088‬‬ ‫‪225.447‬‬ ‫‪227.934‬‬ ‫‪235.210‬‬

‫‪227.346‬‬ ‫‪225.772‬‬ ‫‪234.323‬‬ ‫‪213.619‬‬ ‫‪221.484‬‬ ‫‪236.419‬‬ ‫‪233.918‬‬ ‫‪226.765‬‬ ‫‪228.200‬‬ ‫‪225.799‬‬ ‫‪236.399‬‬ ‫‪222.523‬‬ ‫‪232.976‬‬ ‫‪232.067‬‬ ‫‪230.148‬‬ ‫‪227.077‬‬ ‫‪233.256‬‬ ‫‪231.062‬‬ ‫‪224.050‬‬ ‫‪232.032‬‬ ‫‪233.622‬‬ ‫‪223.599‬‬ ‫‪229.100‬‬ ‫‪242.860‬‬ ‫‪225.123‬‬ ‫‪230.082‬‬ ‫‪228.468‬‬ ‫‪235.623‬‬ ‫‪226.745‬‬ ‫‪229.727‬‬ ‫‪232.891‬‬ ‫‪224.096‬‬

‫)‪z(t‬‬ ‫‪229.574‬‬ ‫‪224.077‬‬ ‫‪230.713‬‬ ‫‪215.405‬‬ ‫‪228.758‬‬ ‫‪234.678‬‬ ‫‪234.155‬‬ ‫‪227.075‬‬ ‫‪230.421‬‬ ‫‪224.663‬‬ ‫‪233.335‬‬ ‫‪223.571‬‬ ‫‪234.561‬‬ ‫‪232.473‬‬ ‫‪231.653‬‬ ‫‪230.146‬‬ ‫‪233.444‬‬ ‫‪230.076‬‬ ‫‪221.171‬‬ ‫‪230.582‬‬ ‫‪232.344‬‬ ‫‪224.880‬‬ ‫‪227.449‬‬ ‫‪239.660‬‬ ‫‪225.860‬‬ ‫‪229.792‬‬ ‫‪235.112‬‬ ‫‪238.292‬‬ ‫‪227.805‬‬ ‫‪230.794‬‬ ‫‪231.092‬‬ ‫‪225.020‬‬

‫اوﻻ ﻧﺮﺳﻢ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻓﻲ ﻣﺨﻄﻂ زﻣﻨﻲ ‪ Time Plot‬ﺑﺈﺳﺘﺨﺪام اﻟﺤﺰﻣﺔ اﻹﺣﺼﺎﺋﻴﺔ ‪MINITAB‬‬ ‫آﺎﻟﺘﺎﻟﻲ‪ ٥٠) :‬ﻣﺸﺎهﺪة ﻓﻘﻂ(‬

‫‪2 4 2‬‬

‫)‪z(t‬‬

‫‪2 3 2‬‬

‫‪2 2 2‬‬ ‫‪5 0‬‬

‫‪4 0‬‬

‫‪3 0‬‬

‫ﺛﺎﻧﻴﺎ ﻧﺤﺴﺐ وﻧﺮﺳﻢ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﻌﻴﻨﻲ‬

‫‪٨٨‬‬

‫‪2 0‬‬

‫‪1 0‬‬

‫‪In d e x‬‬

Autocorrelation

A u t o c o r r e la t io n F u n c t io n f o r z ( t ) 1 .0 0 .8 0 .6 0 .4 0 .2 0 .0 -0 .2 -0 .4 -0 .6 -0 .8 -1 .0

5

Lag 1 2 3 4 5 6 7

C o rr 0 0 0 -0 -0 -0 -0

.8 .5 .1 .1 .3 .3 .2

4 1 5 6 3 3 0

T 13 5 1 -1 -2 -2 -1

.2 .2 .3 .4 .9 .8 .7

7 0 5 6 8 7 3

10

LBQ 1 2 2 2 2 3 3

7 4 4 5 8 1 2

8 4 9 6 3 1 1

.2 .2 .7 .2 .7 .2 .9

1 9 1 1 5 3 5

Lag

1 1 1 1 1

C o rr

8 -0 .0 2 -0 9 0 .1 4 1 0 0 .2 3 1 1 0 .2 3 1 2 0 .1 5 1 3 0 .0 2 0 4 -0 .1 2 -0

T .1 .1 .9 .8 .2 .1 .9

7 8 5 8 2 5 6

15

LBQ 3 3 3 3 3 3 3

2 2 4 5 6 6 6

2 7 1 5 1 1 5

.0 .2 .4 .1 .1 .2 .0

Lag

5 4 8 6 7 5 4

1 1 1 1 1 2

5 6 7 8 9 0

C o rr -0 -0 -0 -0 0 0

.2 .2 .1 .1 .0 .0

1 2 8 0 0 8

20

T -1 -1 -1 -0 0 0

.6 .7 .4 .7 .0 .6

5 9 2 6 2 1

LBQ 3 3 3 4 4 4

7 8 9 0 0 0

6 9 8 1 1 3

.3 .9 .8 .3 .3 .0

2 3 1 7 7 5

‫ﺛﺎﻟﺜﺎ ﻧﺤﺴﺐ وﻧﺮﺳﻢ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻲ‬

Partial Autocorrelation

P a r t ia l A u t o c o r r e la t io n F u n c t io n f o r z ( t ) 1 .0 0 .8 0 .6 0 .4 0 .2 0 .0 -0 .2 -0 .4 -0 .6 -0 .8 -1 .0

5

Lag 1 2 3 4 5 6 7

PAC

10

T

0 .8 4 1 3 .2 7 -0 .6 6 -1 0 .3 9 -1 .1 3 -0 .0 7 -1 .0 0 -0 .0 6 1 .7 8 0 .1 1 2 .1 8 0 .1 4 -0 .2 2 -0 .0 1

Lag

15

PAC

T

8 0 .0 7 9 -0 .0 7 1 0 0 .0 5 1 1 -0 .0 7 1 2 0 .0 6 1 3 -0 .1 5 1 4 0 .0 1

1 .0 6 -1 .1 2 0 .7 3 -1 .1 8 0 .9 0 -2 .4 1 0 .1 8

Lag

20

PAC

T

1 5 0 .0 4 1 6 -0 .0 6 1 7 0 .0 4 1 8 -0 .0 7 1 9 0 .0 9 2 0 -0 .0 9

0 .6 7 -0 .9 3 0 .6 0 -1 .1 4 1 .4 4 -1 .3 9

‫ﻣﻦ أﻧﻤﺎط اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ و اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻲ ﻧﻼﺣﻆ ان اﻟﻤﺘﺴﻠﺴﻠﺔ ﺗﺘﺒﻊ ﻧﻤﻮذج‬ ‫ وﻟﻬﺬا ﻧﻄﺒﻖ اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح ﻋﻠﻲ اﻟﻤﺸﺎهﺪات‬AR(٢) MTB > Name c7 = 'RESI1' MTB > ARIMA 2 0 0 'z(t)' 'RESI1'; SUBC> Constant; SUBC> Forecast 10 c4 c5 c6; SUBC> GACF; SUBC> GPACF; SUBC> GHistogram; SUBC> GNormalplot.

ARIMA Model ARIMA model for z(t) Estimates at each iteration Iteration SSE Parameters 0 4257.23 0.100 0.100 1 3528.31 0.250 0.012 2 2889.23 0.400 -0.076 3 2338.97 0.550 -0.165 4 1877.39 0.700 -0.253 5 1504.46 0.850 -0.342 6 1220.13 1.000 -0.430

٨٩

183.784 169.535 155.360 141.201 127.051 112.913 98.789

7 1024.34 1.150 8 916.97 1.300 9 894.38 1.402 10 894.31 1.408 11 894.31 1.408 Relative change in each estimate

-0.519 84.690 -0.608 70.623 -0.668 61.154 -0.672 60.670 -0.672 60.646 less than 0.0010

Final Estimates of Parameters Type Coef StDev AR 1 1.4079 0.0473 AR 2 -0.6720 0.0474 Constant 60.6458 0.1203 Mean 229.638 0.456

T 29.78 -14.19 504.11

Number of observations: 250 Residuals: SS = 893.567 (backforecasts excluded) MS = 3.618 DF = 247 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 17.5(DF=10) 27.2(DF=22) 49.7(DF=34)

48 67.7(DF=46)

Forecasts from period 250 Period 251 252 253 254 255 256 257 258 259 260

Forecast 224.939 226.747 228.725 230.296 231.177 231.363 231.033 230.442 229.833 229.372

95 Percent Limits Lower Upper 221.211 228.668 220.308 233.186 220.642 236.808 221.546 239.045 222.311 240.044 222.494 240.233 222.070 239.996 221.327 239.558 220.600 239.067 220.090 238.655

zt = 60.6458 + 1.4079 zt −1 − 0.672 zt −2 + at , at

( ) s.e. (φˆ ) = 0.0474,

Actual

:‫وﻧﺴﺘﻨﺘﺞ اﻟﺘﺎﻟﻲ‬ ‫ اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح هﻮ‬-١

WN ( 0,3.618)

:‫ ﻟﻬﺎ هﻲ‬t ‫ اﻟﻤﻌﺎﻟﻢ اﻟﻤﻘﺪرة وإﻧﺤﺮاﻓﻬﺎ اﻟﻤﻌﻴﺎري و ﻗﻴﻤﺔ‬-٢

φˆ1 = 1.4079, s.e. φˆ1 = 0.0473, t = 29.78 φˆ2 = −0.672,

2

t = −14.19

µˆ = 229.638, s.e. ( µˆ ) = 0.456

( )

δˆ = 60.6458, s.e. δˆ = 0.1203, t = 504.11 σˆ 2 = 3.618, with d . f . = 247

:‫راﺑﻌﺎ ﻧﻔﺤﺺ اﻟﺒﻮاﻗﻲ‬ ‫ إﺧﺘﺒﺎر ﻣﺘﻮﺳﻂ اﻟﺒﻮاﻗﻲ‬-١ MTB > ZTest 0.0 3.618 'RESI1'; SUBC> Alternative 0.

Z-Test Test of mu = 0.000 vs mu not = 0.000 The assumed sigma = 3.62 Variable RESI1

N 250

Mean -0.005

StDev 1.894

SE Mean 0.229

٩٠

Z -0.02

P 0.98

‫ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ اﻟﺼﻔﺮﻳﺔ ﺑﺄن اﻟﻤﺘﻮﺳﻂ ﻳﺴﺎوي اﻟﺼﻔﺮ‬ ‫ إﺧﺘﺒﺎر ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬-٢ MTB > Runs 0 'RESI1'.

Runs Test RESI1 K =

0.0000

The observed number of runs = 125 The expected number of runs = 125.8720 129 Observations above K 121 below The test is significant at 0.9119 Cannot reject at alpha = 0.05

‫ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ اﻟﺼﻔﺮﻳﺔ ﺑﺄن اﻟﺒﻮاﻗﻲ ﻋﺸﻮاﺋﻴﺔ‬ ‫ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﺒﻮاﻗﻲ‬-٣ A C F o f R e s id u a ls f o r z ( t) ( w it h 9 5 % c o n f id e n c e l im it s f o r t h e a u t o c o r r e l a t io n s ) 1 .0 0 .8

Autocorrelation

0 .6 0 .4 0 .2 0 .0 - 0 .2 - 0 .4 - 0 .6 - 0 .8 - 1 .0 5

10

15

20

25

30

35

40

45

50

55

60

Lag

‫ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﺒﻮاﻗﻲ‬-٤ P A C F o f R e s id u a ls f o r z (t) ( w it h 9 5 % c o n f id e n c e l im it s f o r t h e p a r t ia l a u t o c o r r e l a t io n s ) 1 .0

Partial Autocorrelation

0 .8 0 .6 0 .4 0 .2 0 .0 - 0 .2 - 0 .4 - 0 .6 - 0 .8 - 1 .0 5

10

15

20

25

30

Lag

٩١

35

40

45

50

55

60

‫ﻧﻼﺣﻆ ان أﻧﻤﺎط اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﺗﺘﺒﻊ أﻧﻤﺎط ﻣﺘﺴﻠﺴﻠﺔ اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‬ ‫‪ -٥‬إﺧﺘﺒﺎر ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ ‪:‬‬ ‫ا‪ -‬ﻧﺮﺳﻢ اﻟﻤﻀﻠﻊ اﻟﺘﻜﺮاري ﻟﻠﺒﻮاﻗﻲ‬ ‫‪H istogram of the R esiduals‬‬ ‫))‪(res p on s e is z(t‬‬ ‫‪30‬‬

‫‪10‬‬

‫‪0‬‬ ‫‪5‬‬

‫‪-5‬‬

‫‪0‬‬

‫‪R es idual‬‬

‫ﻧﻼﺣﻆ أﻧﻪ ﻣﺘﻨﺎﻇﺮ وﻟﺔ ﺷﻜﻞ اﻟﺘﻮزﻳﻊ اﻟﻄﺒﻴﻌﻲ ﺗﻘﺮﻳﺒﺎ‪ .‬وهﺬا ﻻﻳﻜﻔﻲ ﺑﻞ ﻳﺠﺐ ان ﻧﻨﻈﺮ اﻟﻰ‪:‬‬ ‫ب‪ -‬رﺳﻢ اﻻﺣﺘﻤﺎل اﻟﻄﺒﻴﻌﻲ ‪Normal Probability Plot‬‬ ‫‪N orm al P rob ab ility P lot for R E S I1‬‬

‫‪-4 .6 E -0 3‬‬

‫‪M ean:‬‬

‫‪1 .8 9 4 3 6‬‬

‫‪S tD e v:‬‬

‫‪99‬‬

‫‪95‬‬ ‫‪90‬‬ ‫‪80‬‬

‫‪20‬‬ ‫‪10‬‬ ‫‪5‬‬

‫‪1‬‬

‫‪4‬‬

‫‪2‬‬

‫‪0‬‬

‫‪D a ta‬‬

‫واﺿﺢ ﻣﻦ اﻟﺮﺳﻢ أن اﻟﺒﻮاﻗﻲ ﻃﺒﻴﻌﻴﺔ وﻟﻠﺘﺄآﺪ ﻧﻘﻮم‪:‬‬ ‫ج‪ -‬ﺑﺈﺧﺘﺒﺎر ‪ K-S Test‬ﻟﻄﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬

‫‪٩٢‬‬

‫‪-2‬‬

‫‪-4‬‬

‫‪-6‬‬

‫‪Percent‬‬

‫‪70‬‬ ‫‪60‬‬ ‫‪50‬‬ ‫‪40‬‬ ‫‪30‬‬

‫‪Frequency‬‬

‫‪20‬‬

Normal Probability Plot

.999 .99

Probability

.95 .80 .50 .20 .05 .01 .001 -5

0

5

RESI1 Average: -0.0046305 StDev: 1.89436 N: 250

Kolmogorov-Smirnov Normality Test D+: 0.020 D-: 0.029 D : 0.029 Approximate P-Value > 0.15

‫ وهﻲ أآﺒﺮ ﻣﻦ‬٠٫١٥ ‫ ﻟﻺﺧﺘﺒﺎر هﻲ‬P-Value ‫ﺳﻤﻴﺮﻧﻮف اﻋﻄﻰ اﻟـ‬-‫إﺧﺘﺒﺎر آﻮﻟﻤﻮﺟﻮروف‬ .‫ أي اﻧﻨﺎ ﻻﻧﺮﻓﺾ ﻓﺮﺿﻴﺔ ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬α = 0.05 :‫ﺗﻮﻟﻴﺪ ﺗﻨﺒﺆات‬ ‫ ﻗﻴﻢ ﻣﺴﺘﻘﺒﻠﻴﺔ‬١٠ ‫اﺳﺘﺨﺪﻣﻨﺎ اﻟﻨﻤﻮذج ﻟﻠﺘﻨﺒﺆ ﻋﻦ‬ Forecasts from period 250 Period 251 252 253 254 255 256 257 258 259 260

95 Percent Limits Lower Upper 221.211 228.668 220.308 233.186 220.642 236.808 221.546 239.045 222.311 240.044 222.494 240.233 222.070 239.996 221.327 239.558 220.600 239.067 220.090 238.655

Forecast 224.939 226.747 228.725 230.296 231.177 231.363 231.033 230.442 229.833 229.372

Actual

‫ ﻓﺘﺮات ﺗﻨﺒﺆ‬٩٥% ‫واﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻳﻌﻄﻲ اﻟﺘﻨﺒﺆات و‬ Forecast of 10 future values with 95% limits

C4

240

230

220 0

1

2

3

4

5

6

7

8

9

C8

٩٣

10

‫‪Forecast of 10 future values with 95% limits‬‬ ‫‪245‬‬

‫‪235‬‬

‫‪C9‬‬ ‫‪225‬‬

‫‪215‬‬ ‫‪200‬‬

‫‪0‬‬

‫‪100‬‬

‫‪C8‬‬

‫‪Forecast of 10 future values with 95% limits‬‬

‫‪240‬‬

‫‪C9‬‬

‫‪230‬‬

‫‪220‬‬ ‫‪50‬‬

‫‪60‬‬

‫‪30‬‬

‫‪40‬‬

‫‪20‬‬

‫‪10‬‬

‫‪0‬‬

‫‪C8‬‬

‫اﻟﺸﻜﻞ اﻷول ﻳﺒﻴﻦ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ ﺑﻜﺎﻣﻠﻬﺎ ﻣﻊ اﻟﺘﻨﺒﺆات واﻟﺸﻜﻞ اﻟﺜﺎﻧﻲ ﻟﻠﺨﻤﺴﻴﻦ ﻗﻴﻤﺔ اﻷﺧﻴﺮة ﻣﻊ‬ ‫اﻟﺘﻨﺒﺆات ﻟﺘﻮﺿﻴﺢ ﺷﻜﻞ داﻟﺔ اﻟﺘﻨﺒﺆ‪.‬‬ ‫ﺣﺎﻟﺔ دراﺳﺔ‪:‬‬ ‫اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺘﺎﻟﻴﺔ اﻟﻤﺘﻮﺳﻂ اﻟﻴﻮﻣﻲ ﻟﻌﺪد اﻟﺘﻠﻔﺰﻳﻮﻧﺎت اﻟﻤﻌﻴﺒﺔ ﻓﻲ ﺧﻂ إﻧﺘﺎج ﻣﺼﻨﻊ ﻣﺎ )إﻗﺮأ ﺳﻄﺮا‬ ‫ﺑﺴﻄﺮ(‬ ‫‪2.09‬‬ ‫‪1.57‬‬ ‫‪2.07‬‬ ‫‪1.78‬‬

‫‪2.00‬‬ ‫‪1.42‬‬ ‫‪1.82‬‬ ‫‪1.68‬‬

‫‪1.76‬‬ ‫‪1.54‬‬ ‫‪1.85‬‬ ‫‪1.79‬‬

‫‪2.83‬‬ ‫‪1.46‬‬ ‫‪2.08‬‬ ‫‪1.37‬‬

‫‪3.44‬‬ ‫‪2.05‬‬ ‫‪1.42‬‬ ‫‪1.15‬‬

‫‪2.40‬‬ ‫‪2.50‬‬ ‫‪1.39‬‬ ‫‪1.25‬‬

‫اﻟﻤﺨﻄﻂ اﻟﺰﻣﻨﻲ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ‬

‫‪٩٤‬‬

‫‪1.95‬‬ ‫‪2.25‬‬ ‫‪1.18‬‬ ‫‪1.61‬‬

‫‪2.70‬‬ ‫‪1.58‬‬ ‫‪1.27‬‬ ‫‪1.77‬‬

‫‪1.54‬‬ ‫‪1.25‬‬ ‫‪1.08‬‬ ‫‪2.91‬‬

‫‪Defects‬‬ ‫‪1.20 1.50‬‬ ‫‪1.89 1.80‬‬ ‫‪1.40 1.51‬‬ ‫‪2.32 1.23‬‬ ‫‪1.84‬‬

3 .5

3 .0

Defects

2 .5 2 .0 1 .5 1 .0

In d e x

10

20

30

40

‫اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬ Autocorrelation

Autocorrelation Function for Defects 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

1

2

3

Lag

Corr

4

5

6

7

8

T

LBQ

Lag

Corr

T

LBQ

1 0.43 2.88 2 0.26 1.49 3 0.14 0.77 4 0.08 0.43 5 -0.09 -0.46 6 -0.07 -0.39 7 -0.21 -1.10

8.84 12.18 13.18 13.50 13.89 14.18 16.57

8 9 10 11

-0.11 -0.05 -0.01 -0.04

-0.57 -0.27 -0.04 -0.19

17.25 17.41 17.41 17.50

9

10

11

Partial Autocorrelation

Partial Autocorrelation Function for Defects 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

1

2

3

Lag PAC

4

5

T

1 0.43 2.88 2 0.09 0.63 3 -0.00 -0.01 4 0.00 0.00 5 -0.16 -1.07 6 0.00 0.02 7 -0.18 -1.19

6

Lag PAC

7

8

T

8 0.07 0.44 9 0.05 0.35 10 0.01 0.09 11 -0.03 -0.23

٩٥

9

10

11

:‫ واﻟﺬي ﻳﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬AIC ‫ﻹﺧﺘﻴﺎر اﻟﻨﻤﻮذج اﻟﻤﻨﺎﺳﺐ ﺳﻮف ﻧﺴﺘﺨﺪم ﻣﻌﻴﺎر اﻹﻋﻼم اﻟﺬاﺗﻲ‬ AIC ( m ) = n ln σ a2 + 2m ‫ ﻋﺪد اﻟﻤﻌﺎﻟﻢ اﻟﻤﻘﺪرة ﻓﻲ اﻟﻨﻤﻮذج وﻧﺨﺘﺎر اﻟﻨﻤﻮذج اﻟﺬي ﻳﻌﻄﻲ‬m ‫ﺣﻴﺚ‬ min AIC ( m ) m

:‫ﺳﻮف ﻧﻄﺒﻖ اﻟﻨﻤﺎذج ﻋﻠﻲ اﻟﺘﻮاﻟﻲ‬ MTB > ARIMA 1 0 0 'Defects' 'RESI1'; SUBC> Constant;

ARIMA Model ARIMA model for Defects Final Estimates of Parameters Type Coef StDev AR 1 0.4421 0.1365 Constant 0.99280 0.06999 Mean 1.7795 0.1254

T 3.24 14.19

Number of observations: 45 Residuals: SS = 9.47811 (backforecasts excluded) MS = 0.22042 DF = 43 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 4.9(DF=11) 8.9(DF=23) 30.9(DF=35)

48 * (DF= *)

MTB > ARIMA 2 0 0 'Defects' 'RESI2'; SUBC> Constant;

ARIMA Model ARIMA model for Defects Final Estimates of Parameters Type Coef StDev AR 1 0.3999 0.1533 AR 2 0.0989 0.1531 Constant 0.89019 0.07047 Mean 1.7762 0.1406

T 2.61 0.65 12.63

Number of observations: 45 Residuals: SS = 9.38567 (backforecasts excluded) MS = 0.22347 DF = 42 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 4.0(DF=10) 8.1(DF=22) 28.8(DF=34) MTB > ARIMA 1 0 1 'Defects' 'RESI3'; SUBC> Constant;

ARIMA Model ARIMA model for Defects Final Estimates of Parameters Type Coef StDev AR 1 0.5983 0.2691 MA 1 0.1926 0.3294 Constant 0.71334 0.05693

T 2.22 0.58 12.53

٩٦

48 * (DF= *)

Mean

1.7759

0.1417

Number of observations: 45 Residuals: SS = 9.39423 (backforecasts excluded) MS = 0.22367 DF = 42 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 4.1(DF=10) 8.3(DF=22) 29.1(DF=34)

48 * (DF= *)

MTB > ARIMA 0 0 1 'Defects' 'RESI4'; SUBC> Constant;

ARIMA Model ARIMA model for Defects Final Estimates of Parameters Type Coef StDev MA 1 -0.3409 0.1431 Constant 1.78480 0.09651 Mean 1.78480 0.09651

T -2.38 18.49

Number of observations: 45 Residuals: SS = 10.0362 (backforecasts excluded) MS = 0.2334 DF = 43 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 8.0(DF=11) 13.2(DF=23) 35.7(DF=35)

48 * (DF= *)

MTB > ARIMA 0 0 2 'Defects' 'RESI5'; SUBC> Constant;

ARIMA Model ARIMA model for Defects Final Estimates of Parameters Type Coef StDev MA 1 -0.3869 0.1516 MA 2 -0.1816 0.1516 Constant 1.7839 0.1118 Mean 1.7839 0.1118

T -2.55 -1.20 15.96

Number of observations: 45 Residuals: SS = 9.61059 (backforecasts excluded) MS = 0.22882 DF = 42 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 4.6(DF=10) 9.2(DF=22) 31.0(DF=34) MTB > ARIMA 2 0 1 'Defects' 'RESI6'; SUBC> Constant;

ARIMA Model ARIMA model for Defects Final Estimates of Parameters Type Coef StDev AR 1 0.4134 1.5680

T 0.26

٩٧

48 * (DF= *)

AR 2 MA 1 Constant Mean

0.0929 0.0136 0.87675 1.7761

0.7113 1.5749 0.07036 0.1425

0.13 0.01 12.46

Number of observations: 45 Residuals: SS = 9.38561 (backforecasts excluded) MS = 0.22892 DF = 41 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 4.0(DF= 9) 8.1(DF=21) 28.8(DF=33)

48 * (DF= *)

MTB > ARIMA 1 0 2 'Defects' 'RESI7'; SUBC> Constant;

ARIMA Model ARIMA model for Defects * ERROR * Model cannot be estimated with these data MTB > ARIMA 2 0 2 'Defects' 'RESI8'; SUBC> Constant;

ARIMA Model ARIMA model for Defects Final Estimates of Parameters Type Coef StDev AR 1 1.6720 0.1165 AR 2 -0.7263 0.1251 MA 1 1.3199 0.0184 MA 2 -0.3196 0.0731 Constant 0.096224 0.003323 Mean 1.77238 0.06121

T 14.35 -5.80 71.63 -4.37 28.95

Number of observations: 45 Residuals: SS = 8.33225 (backforecasts excluded) MS = 0.20831 DF = 40 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 4.7(DF= 8) 8.9(DF=20) 29.7(DF=32)

48 * (DF= *)

:‫وﻧﻠﺨﺺ ذﻟﻚ ﺑﺎﻟﺠﺪول اﻟﺘﺎﻟﻲ‬ Model __________ AR (1) AR ( 2 ) MA (1) MA ( 2 ) ARMA (1,1) ARMA ( 2,1) ARMA (1, 2 ) ARMA ( 2, 2 )

σˆ

2

________ 0.22042 0.22347 0.23340 0.22882 0.22367 0.22892 − 0.20831

m ___ 3 4 3 4 4 5 − 6

AIC _________ −62.0499 −59.4315 −59.4751 −58.3669 −59.3913 −56.3472 − −58.5928

٩٨

‫‪min AIC ( m ) = −62.0499‬‬ ‫‪m‬‬

‫أي ان أﻓﻀﻞ ﻧﻤﻮذج هﻮ )‪. AR(١‬‬ ‫ﻳﺘﺮك ﻟﻠﻄﺎﻟﺐ آﺘﻤﺮﻳﻦ ﻓﺤﺺ اﻟﺒﻮاﻗﻲ وﺗﻮﻟﻴﺪ ﺗﻨﺒﺆات‪.‬‬

‫ﺣﺎﻟﺔ دراﺳﺔ‪:‬‬ ‫اﻟﻤﺘﺴﻠﺴﺔ اﻟﺘﺎﻟﻴﺔ هﻲ دﺧﻞ اﻟﻤﺒﻴﻌﺎت اﻟﺴﻨﻮﻳﺔ ﺑﻤﻼﻳﻴﻦ اﻟﺮﻳﺎﻻت ﻟﺸﺮآﺔ ﻣﺎ‬ ‫‪5.43‬‬ ‫‪3.88‬‬ ‫‪3.57‬‬ ‫‪2.75‬‬ ‫‪6.06‬‬ ‫‪5.80‬‬ ‫‪5.16‬‬ ‫‪6.64‬‬ ‫‪6.43‬‬ ‫‪7.86‬‬

‫‪3.80‬‬ ‫‪4.30‬‬ ‫‪3.45‬‬ ‫‪4.80‬‬ ‫‪6.12‬‬ ‫‪6.08‬‬ ‫‪5.71‬‬ ‫‪7.49‬‬ ‫‪7.53‬‬ ‫‪7.50‬‬

‫‪4.14‬‬ ‫‪5.42‬‬ ‫‪1.22‬‬ ‫‪3.08‬‬ ‫‪5.65‬‬ ‫‪4.78‬‬ ‫‪5.61‬‬ ‫‪6.09‬‬ ‫‪5.62‬‬ ‫‪8.27‬‬

‫‪4.60‬‬ ‫‪3.91‬‬ ‫‪3.98‬‬ ‫‪5.43‬‬ ‫‪5.52‬‬ ‫‪5.67‬‬ ‫‪5.63‬‬ ‫‪6.64‬‬ ‫‪7.59‬‬ ‫‪8.75‬‬

‫‪4.77‬‬ ‫‪5.07‬‬ ‫‪2.65‬‬ ‫‪4.40‬‬ ‫‪4.79‬‬ ‫‪4.89‬‬ ‫‪5.70‬‬ ‫‪4.72‬‬ ‫‪7.27‬‬ ‫‪8.50‬‬

‫‪3.45‬‬ ‫‪3.78‬‬ ‫‪3.28‬‬ ‫‪3.84‬‬ ‫‪6.11‬‬ ‫‪4.99‬‬ ‫‪5.75‬‬ ‫‪6.08‬‬ ‫‪9.01‬‬ ‫‪7.23‬‬

‫‪3.99‬‬ ‫‪6.16‬‬ ‫‪4.05‬‬ ‫‪5.14‬‬ ‫‪6.46‬‬ ‫‪6.12‬‬ ‫‪6.36‬‬ ‫‪6.57‬‬ ‫‪7.49‬‬ ‫‪7.53‬‬

‫‪5.51‬‬ ‫‪4.05‬‬ ‫‪4.08‬‬ ‫‪4.00‬‬ ‫‪4.31‬‬ ‫‪6.23‬‬ ‫‪8.02‬‬ ‫‪7.56‬‬ ‫‪6.69‬‬ ‫‪7.67‬‬

‫‪5.74‬‬ ‫‪2.54‬‬ ‫‪4.61‬‬ ‫‪1.58‬‬ ‫‪4.99‬‬ ‫‪5.05‬‬ ‫‪7.07‬‬ ‫‪6.87‬‬ ‫‪7.22‬‬ ‫‪8.22‬‬

‫ﻣﺨﻄﻂ زﻣﻨﻲ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ‬

‫‪9‬‬ ‫‪8‬‬ ‫‪7‬‬ ‫‪6‬‬

‫‪Sales‬‬

‫‪5‬‬ ‫‪4‬‬ ‫‪3‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫‪100‬‬

‫‪90‬‬

‫‪80‬‬

‫‪70‬‬

‫‪60‬‬

‫‪50‬‬

‫دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ اﻟﻌﻴﻨﻴﺔ‬

‫‪٩٩‬‬

‫‪40‬‬

‫‪30‬‬

‫‪20‬‬

‫‪10‬‬

‫‪In d e x‬‬

‫‪Sales‬‬ ‫‪3.49‬‬ ‫‪3.96‬‬ ‫‪2.89‬‬ ‫‪2.52‬‬ ‫‪5.77‬‬ ‫‪3.20‬‬ ‫‪5.13‬‬ ‫‪7.20‬‬ ‫‪7.26‬‬ ‫‪6.42‬‬

Autocorrelation

Autocorrelation Function for Sales 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

5

Lag Corr 1 2 3 4 5 6 7

0.71 0.60 0.65 0.64 0.59 0.59 0.51

T

LBQ

7.10 51.97 4.26 89.85 3.94 134.64 3.36 177.75 2.83 215.72 2.63 253.97 2.12 282.61

15

Lag Corr 8 9 10 11 12 13 14

0.56 0.49 0.49 0.51 0.42 0.38 0.45

T

LBQ

2.22 317.32 1.87 344.49 1.79 371.67 1.82 401.71 1.46 422.60 1.29 439.91 1.50 464.06

Lag Corr 15 16 17 18 19 20 21

0.41 0.35 0.31 0.30 0.36 0.31 0.26

T

25

LBQ

1.32 483.85 1.13 499.04 0.97 510.68 0.92 521.52 1.11 537.81 0.95 550.11 0.77 558.52

Lag Corr 22 23 24 25

0.22 0.17 0.21 0.25

T

LBQ

0.67 565.04 0.50 568.77 0.64 574.90 0.75 583.43

1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

5

Lag PAC

T

1 0.71 7.10 2 0.20 1.99 3 0.35 3.53 4 0.15 1.49 5 0.09 0.92 6 0.10 1.03 7 -0.13 -1.27

15

Lag PAC

25

T

Lag PAC

T

8 0.19 1.93 9 -0.17 -1.70 10 0.14 1.44 11 0.03 0.34 12 -0.15 -1.52 13 0.02 0.25 14 0.04 0.39

15 0.03 16 -0.10 17 -0.08 18 -0.01 19 0.14 20 -0.04 21 0.00

0.32 -0.96 -0.84 -0.10 1.43 -0.44 0.04

Lag PAC

T

22 -0.17 -1.68 23 -0.10 -1.00 24 0.15 1.55 25 0.03 0.34

.‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﻌﻴﻨﻲ ﺗﺪل ﻋﻠﻰ ﺗﺨﺎﻣﺪ ﺑﻄﻴﺊ ﻣﻤﺎ ﻗﺪ ﻳﺪل ﻋﻠﻰ ﻋﺪم إﺳﺘﻘﺮار ﻓﻲ اﻟﻤﺘﻮﺳﻂ‬ ‫ وﻧﺮﺳﻤﻬﺎ‬wt = ∇zt ‫ﻟﻨﺠﺮب اﻟﺘﻔﺮﻳﻖ اﻷول ﻟﻠﻤﺘﺴﻠﺴﻠﺔ‬

3 2 1

w(t)

Partial Autocorrelation

Partial Autocorrelation Function for Sales

0 -1 -2 -3

In d e x

10

20

30

40

50

60

70

80

90

100

‫ دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﺴﺘﻘﺮة‬.‫ﺗﺒﺪو اﻟﻤﺘﺴﻠﺴﻠﺔ ﻣﺴﺘﻘﺮة اﻵن‬

١٠٠

Autocorrelation

Autocorrelation Function for w(t) 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

2

Lag Corr

12

T

LBQ

1 -0.30 -3.00 2 -0.31 -2.86 3 0.11 0.90 4 0.04 0.31 5 -0.04 -0.33 6 0.14 1.15 7 -0.24 -1.98

9.26 19.32 20.49 20.63 20.80 22.80 29.02

Lag Corr

T

LBQ

8 0.19 1.53 9 -0.07 -0.53 10 -0.07 -0.52 11 0.16 1.27 12 -0.07 -0.53 13 -0.14 -1.04 14 0.15 1.10

33.07 33.58 34.08 37.12 37.67 39.85 42.37

Lag Corr

22

T

LBQ

15 0.00 0.00 16 0.02 0.13 17 -0.07 -0.52 18 -0.15 -1.09 19 0.20 1.48 20 0.01 0.05 21 -0.04 -0.25

42.37 42.40 42.99 45.65 50.73 50.74 50.90

Lag Corr

T

LBQ

22 0.05 0.35 51.22 23 -0.18 -1.28 55.44 24 0.03 0.20 55.55

Partial Autocorrelation

Partial Autocorrelation Function for w(t) 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

2

12

22

Lag PAC

T

Lag PAC

T

Lag PAC

T

1 -0.30 2 -0.44 3 -0.22 4 -0.21 5 -0.17 6 0.06 7 -0.26

-3.00 -4.41 -2.23 -2.05 -1.72 0.59 -2.55

8 0.12 9 -0.16 10 -0.06 11 0.09 12 -0.06 13 -0.02 14 -0.07

1.17 -1.59 -0.60 0.93 -0.59 -0.17 -0.71

15 0.06 16 0.07 17 -0.03 18 -0.17 19 -0.03 20 -0.10 21 0.09

0.59 0.70 -0.32 -1.65 -0.31 -0.95 0.85

Lag PAC

T

22 0.08 0.80 23 -0.15 -1.53 24 -0.05 -0.55

:‫ واﻟﺬي ﻳﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬AIC ‫ﻹﺧﺘﻴﺎر اﻟﻨﻤﻮذج اﻟﻤﻨﺎﺳﺐ ﺳﻮف ﻧﺴﺘﺨﺪم ﻣﻌﻴﺎر اﻹﻋﻼم اﻟﺬاﺗﻲ‬ AIC ( m ) = n ln σ a2 + 2m ‫ ﻋﺪد اﻟﻤﻌﺎﻟﻢ اﻟﻤﻘﺪرة ﻓﻲ اﻟﻨﻤﻮذج وﻧﺨﺘﺎر اﻟﻨﻤﻮذج اﻟﺬي ﻳﻌﻄﻲ‬m ‫ﺣﻴﺚ‬ min AIC ( m ) m

:‫ﺳﻮف ﻧﻄﺒﻖ اﻟﻨﻤﺎذج ﻋﻠﻲ اﻟﺘﻮاﻟﻲ‬ MTB > ARIMA 1 1 0 'Sales'; SUBC> NoConstant.

ARIMA Model ARIMA model for Sales Final Estimates of Parameters Type Coef StDev AR 1 -0.3114 0.0959

T -3.25

Differencing: 1 regular difference Number of observations: Original series 100, after differencing 99 Residuals: SS = 133.134 (backforecasts excluded)

١٠١

MS =

1.359

DF = 98

Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 31.9(DF=11) 51.2(DF=23) 62.8(DF=35)

48 81.0(DF=47)

MTB > ARIMA 2 1 0 'Sales'; SUBC> NoConstant.

ARIMA Model ARIMA model for Sales Final Estimates of Parameters Type Coef StDev AR 1 -0.4532 0.0897 AR 2 -0.4656 0.0901

T -5.05 -5.17

Differencing: 1 regular difference Number of observations: Original series 100, after differencing 99 Residuals: SS = 104.715 (backforecasts excluded) MS = 1.080 DF = 97 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 21.8(DF=10) 40.9(DF=22) 49.4(DF=34)

48 59.9(DF=46)

MTB > ARIMA 0 1 1 'Sales'; SUBC> NoConstant.

ARIMA Model ARIMA model for Sales Final Estimates of Parameters Type Coef StDev MA 1 0.7636 0.0648

T 11.78

Differencing: 1 regular difference Number of observations: Original series 100, after differencing 99 Residuals: SS = 101.411 (backforecasts excluded) MS = 1.035 DF = 98 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 12.6(DF=11) 27.8(DF=23) 35.9(DF=35)

48 48.5(DF=47)

MTB > ARIMA 0 1 2 'Sales'; SUBC> NoConstant.

ARIMA Model ARIMA model for Sales Final Estimates of Parameters Type Coef StDev MA 1 0.5756 0.0990 MA 2 0.2029 0.0998

T 5.81 2.03

Differencing: 1 regular difference Number of observations: Original series 100, after differencing 99 Residuals: SS = 99.2463 (backforecasts excluded) MS = 1.0232 DF = 97

١٠٢

Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 14.3(DF=10) 28.3(DF=22) 36.5(DF=34)

48 47.0(DF=46)

MTB > ARIMA 1 1 1 'Sales'; SUBC> NoConstant.

ARIMA Model ARIMA model for Sales Final Estimates of Parameters Type Coef StDev AR 1 0.1283 0.1334 MA 1 0.8027 0.0799

T 0.96 10.04

Differencing: 1 regular difference Number of observations: Original series 100, after differencing 99 Residuals: SS = 100.421 (backforecasts excluded) MS = 1.035 DF = 97 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 13.3(DF=10) 27.9(DF=22) 36.1(DF=34)

48 48.2(DF=46)

MTB > ARIMA 2 1 1 'Sales'; SUBC> NoConstant.

ARIMA Model ARIMA model for Sales * WARNING * Back forecasts not dying out rapidly Final Estimates of Parameters Type Coef StDev AR 1 -1.1389 0.0987 AR 2 -0.1440 0.0983 MA 1 -0.9889 0.0002

T -11.53 -1.47 -3987.49

Differencing: 1 regular difference Number of observations: Original series 100, after differencing 99 Residuals: SS = 134.250 (backforecasts excluded) MS = 1.398 DF = 96 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 35.1(DF= 9) 53.5(DF=21) 66.6(DF=33) MTB > ARIMA 1 1 2 'Sales'; SUBC> NoConstant.

ARIMA Model ARIMA model for Sales Final Estimates of Parameters Type Coef StDev AR 1 -0.3476 0.4077 MA 1 0.2422 0.3771 MA 2 0.4506 0.2656

T -0.85 0.64 1.70

Differencing: 1 regular difference

١٠٣

48 83.2(DF=45)

Number of observations: Original series 100, after differencing 99 Residuals: SS = 97.2357 (backforecasts excluded) MS = 1.0129 DF = 96 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 11.9(DF= 9) 25.0(DF=21) 32.1(DF=33)

48 41.8(DF=45)

MTB > ARIMA 2 1 2 'Sales'; SUBC> NoConstant.

ARIMA Model ARIMA model for Sales Final Estimates of Parameters Type Coef StDev AR 1 -0.0691 0.3618 AR 2 -0.2941 0.1450 MA 1 0.5637 0.3737 MA 2 0.0840 0.3266

T -0.19 -2.03 1.51 0.26

Differencing: 1 regular difference Number of observations: Original series 100, after differencing 99 Residuals: SS = 93.6368 (backforecasts excluded) MS = 0.9857 DF = 95 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 11.0(DF= 8) 23.4(DF=20) 30.1(DF=32)

48 36.5(DF=44)

:‫وﻧﻠﺨﺺ ذﻟﻚ ﺑﺎﻟﺠﺪول اﻟﺘﺎﻟﻲ‬ Model σˆ __________ ________ ARI (1,1) 1.359 ARI ( 2,1) 1.080 IMA (1,1) 1.035 IMA (1, 2 ) 1.023 ARIMA (1,1,1) 1.035 ARIMA ( 2,1,1) 1.398 ARIMA (1,1, 2 ) 1.013 ARIMA ( 2,1, 2 ) 0.986 2

m ___ 2 3 2 3 3 4 4 5

AIC _________ 34.368 13.619 7.4057 8.2706 9.4057 41.169 9.2689 8.5741

min AIC ( m ) = 7.406 m

.‫ ﻳﺘﺮك ﻟﻠﻄﺎﻟﺐ آﺘﻤﺮﻳﻦ ﻓﺤﺺ اﻟﺒﻮاﻗﻲ وﺗﻮﻟﻴﺪ ﺗﻨﺒﺆات‬. IMA(١،١) ‫أي ان أﻓﻀﻞ ﻧﻤﻮذج هﻮ‬

١٠٤

١٠٥

‫اﻟﻔﺼﻞ اﻟﺴﺎدس‬ ‫ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ اﻟﺘﻜﺎﻣﻠﻲ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك اﻟﻤﻮﺳﻤﻴﺔ‬

‫‪Seasonal Autoregressive Integrated Moving‬‬ ‫‪Average Models‬‬ ‫اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ ﺗﻌﻄﻲ اﻧﻤﺎط ﻣﺘﺸﺎﺑﻬﺔ ﺗﺘﻜﺮر ﻋﻠﻰ ﻓﺘﺮات زﻣﻨﻴﺔ ﻣﺘﺴﺎوﻳﺔ اﻟﺒﻌﺪ ﻣﺜﻞ‬ ‫ان ﻳﺘﻜﺮر اﻟﻨﻤﻂ آﻞ ارﺑﻌﺔ وﻋﺸﺮون ﺳﺎﻋﺔ او آﻞ ﺳﺒﻌﺔ اﻳﺎم او آﻞ ﺷﻬﺮ او ﺛﻼﺛﺔ اﺷﻬﺮ او ﺳﻨﺔ‪.‬‬ ‫اﻷﺷﻜﺎل اﻟﺘﺎﻟﻴﺔ ﺗﺒﻴﻦ ﻣﺜﻞ هﺬﻩ اﻟﻤﺘﺴﻠﺴﻼت‬ ‫‪S e a s o n a l T im e S e r ie s‬‬

‫‪70‬‬

‫)‪z(t‬‬ ‫‪60‬‬

‫‪50‬‬

‫‪150‬‬

‫‪50‬‬

‫‪100‬‬

‫‪In d e x‬‬

‫‪S e a s o n a l T im e S e r ie s‬‬ ‫‪1000‬‬

‫‪900‬‬

‫)‪z(t‬‬

‫‪800‬‬

‫‪700‬‬

‫‪600‬‬

‫‪150‬‬

‫‪100‬‬

‫‪In d e x‬‬

‫‪50‬‬

‫ﻓﻲ هﺬا اﻟﻔﺼﻞ ﺳﻮف ﻧﺴﺘﻌﺮض ﺧﻮاص اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ وﻃﺮق ﻧﻤﺬﺟﺘﻬﺎ ﺑﻮاﺳﻄﺔ‬ ‫ﻧﻤﺎذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ اﻟﺘﻜﺎﻣﻠﻲ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ‪ SARIMA(p,d,q)(P,D,Q)s‬ﻓﻤﺜﻼ‬ ‫اﻟﻨﻤﻮذج ‪ SARIMA(٠،١،١)(١،١،٠)١٢‬ﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬ ‫) ‪WN ( 0, σ 2‬‬

‫‪١٠٦‬‬

‫‪at‬‬

‫‪(1 − Φ B ) (1 − B ) z = (1 − θ B ) a ,‬‬ ‫‪s‬‬

‫‪t‬‬

‫‪1‬‬

‫‪t‬‬

‫‪1‬‬

‫وﺑﺸﻜﻞ ﻋﺎم ﻓﺈن ﻧﻤﻮذج اﻹﻧﺤﺪار اﻟﺬاﺗﻲ اﻟﺘﻜﺎﻣﻠﻲ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﺑﺎﻟﺪرﺟﺔ ‪(p,d,q)(P,D,Q)s‬‬ ‫‪ SARIMA(p,d,q)(P,D,Q)s‬ﻳﻜﺘﺐ ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫) ‪φ p ( B ) Φ P ( B s ) (1 − B ) (1 − B s ) zt = δ + θ q ( B ) ΘQ ( B s ) at , at WN ( 0, σ 2‬‬ ‫‪D‬‬

‫‪d‬‬

‫ﺣﻴﺚ ) ‪ φ p ( B‬و ) ‪ θ q ( B‬ﻋﻤﺎل اﻹﻧﺤﺪار اﻟﺬاﺗﻲ واﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻏﻴﺮ اﻟﻤﻮﺳﻤﻴﺔ واﻟﺘﻲ ﻣﺮت‬ ‫ﻋﻠﻴﻨﺎ ﺳﺎﺑﻘﺎ و ‪ Φ P ( B s ) = 1 + Φ1B s + Φ 2 B 2 s + L + Φ P B Ps‬ﻋﺎﻣﻞ اﻹﻧﺤﺪار اﻟﺬاﺗﻲ اﻟﻤﻮﺳﻤﻲ و‬ ‫‪ ΘQ ( B s ) = 1 + Θ1B s + Θ2 B 2 s + L + ΘQ B Qs‬ﻋﺎﻣﻞ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك اﻟﻤﻮﺳﻤﻲ وﻳﺴﻤﻰ هﺬا‬ ‫ﺑﺎﻟﻨﻤﻮذج اﻟﻤﻮﺳﻤﻲ اﻟﺘﻀﺎﻋﻔﻲ ‪.Multiplicative Seasonal Models‬‬ ‫ﻣﻼﺣﻈﺔ‪ :‬ﻓﻲ ﺟﻤﻴﻊ اﻟﻨﻤﺎذج اﻟﻘﺎدﻣﺔ ﺳﻴﻜﻮن ﻣﻔﻬﻮﻣﺎ ﺿﻤﻨﻴﺎ أن ) ‪at WN ( 0, σ 2‬‬

‫دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﺒﻌﺾ اﻟﻨﻤﺎذج اﻟﻤﻮﺳﻤﻴﺔ‪:‬‬ ‫‪D‬‬ ‫‪d‬‬ ‫ﻓﻲ اﻹﺷﺘﻘﺎﻗﺎت اﻟﺘﺎﻟﻴﺔ ﺳﻮف ﻧﺴﺘﺨﺪم اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﻮﺳﻤﻴﺔ اﻟﻤﺴﺘﻘﺮة ‪wt = (1 − B ) (1 − B s ) zt‬‬ ‫واﻟﺘﻲ ﺗﺘﺒﻊ اﻟﻨﻤﻮذج ‪SARMA(p,q)(P,Q)s‬‬ ‫) ‪WN ( 0, σ 2‬‬

‫‪φ p ( B ) Φ P ( B ) wt = δ + θ q ( B ) ΘQ ( B ) at , at‬‬ ‫‪s‬‬

‫‪s‬‬

‫ﺳﻮف ﻧﺸﺘﻖ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﻨﻤﻮذج اﻟﻤﻮﺳﻤﻲ اﻟﺘﻀﺎﻋﻔﻲ ‪SARMA(٠،١)(١،١)١٢‬‬ ‫) ‪WN ( 0, σ 2‬‬

‫‪wt = Φ wt −12 + at − θ at −1 − Θat −12 + θ Θat −13 , at‬‬

‫ﺑﻀﺮب ﻃﺮﻓﻲ اﻟﻤﻌﺎدﻟﺔ اﻟﻤﻌﺮﻓﺔ ﻓﻲ ‪ wt‬وأﺧﺬ اﻟﺘﻮﻗﻊ ﻧﺠﺪ‬ ‫‪γ 0 = Φ γ 12 + σ + θ σ − Θ ( Φ − Θ ) σ 2 + θ Θ ( −Φ θ + θ Θ ) σ 2‬‬ ‫‪2‬‬

‫‪2‬‬

‫‪2‬‬

‫⎦⎤) ‪=Φ γ 12 + σ 2 ⎡⎣(1 + θ 2 ) + Θ ( Φ − Θ ) (1 + θ 2‬‬ ‫⎦⎤) ‪=Φ γ 12 + σ 2 (1 + θ 2 ) ⎡⎣1 + Θ ( Φ − Θ‬‬

‫وﺑﻀﺮب ﻃﺮﻓﻲ اﻟﻤﻌﺎدﻟﺔ اﻟﻤﻌﺮﻓﺔ ﻓﻲ ‪ wt −12‬وأﺧﺬ اﻟﺘﻮﻗﻊ ﻧﺠﺪ‬ ‫‪2‬‬

‫‪γ 12 = Φ γ 0 − Θσ + θ Θ ( −θ ) σ‬‬ ‫‪2‬‬

‫) ‪=Φ γ 0 − Θσ 2 (1 + θ 2‬‬

‫وﺑﺤﻞ اﻟﻌﻼﻗﺘﻴﻦ اﻟﺴﺎﺑﻘﺘﻴﻦ ﻧﺠﺪ‬ ‫‪1 + Θ − 2ΦΘ‬‬ ‫‪1 − Φ2‬‬ ‫‪2‬‬ ‫⎡‬ ‫⎤ ) ‪Φ (Θ − Φ‬‬ ‫‪2‬‬ ‫‪2‬‬ ‫‪= σ (1 + θ ) ⎢Φ − Θ +‬‬ ‫⎥‬ ‫⎦⎥ ‪1 − Φ 2‬‬ ‫⎣⎢‬ ‫‪2‬‬

‫أﻳﻀﺎ‬

‫) ‪γ 0 = σ 2 (1 + θ 2‬‬ ‫‪γ 12‬‬

‫) ‪γ 1 = E ( wt wt −1‬‬ ‫) ‪=Φ γ 11 − θσ 2 − ΘE ( at −12 wt −1 ) + θ ΘE ( at −13wt −1‬‬ ‫‪=Φ γ 11 − θσ 2 + θ Θ ( Φ − Θ ) σ 2‬‬

‫و‬ ‫‪2‬‬

‫وﺑﺤﻞ اﻟﻌﻼﻗﺘﻴﻦ اﻟﺴﺎﺑﻘﺘﻴﻦ ﻧﺠﺪ‬

‫‪١٠٧‬‬

‫‪γ 11 = E ( wt wt −11 ) = Φγ 1 + Θθσ‬‬

⎡ ( Θ − Φ )2 ⎤ γ 1 = −θσ ⎢1 + ⎥ 1 − Φ 2 ⎦⎥ ⎣⎢ 2



γ 11 = θσ 2 ⎢ Θ − Φ − ⎣⎢

2 Φ (Θ − Φ ) ⎤ ⎥ 1 − Φ 2 ⎦⎥

‫وﺑﻨﻔﺲ اﻟﻄﺮﻳﻘﺔ ﻳﻤﻜﻦ إﺛﺒﺎت أن‬

γ 2 = γ 3 = L = γ 10 = 0 γ 13 = γ 11 γ k = Φγ k −12 , k > 13

‫وﻣﻦ ﺟﻤﻴﻊ اﻟﻌﻼﻗﺎت اﻟﺴﺎﺑﻘﺔ ﻧﻮﺟﺪ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ‬ k =0 ⎧ 1, ⎪ θ ⎪ − , k =1 2 ⎪ 1+θ ⎪ 0, k = 2,...,10 ⎪ γ ⎪ θ ( Θ − Φ )(1 − ΦΘ ) ρk = k = ⎨ , k = 11 γ 0 ⎪1 + θ 2 1 + Θ2 − 2ΦΘ ⎪ ( Θ − Φ )(1 − ΦΘ ) k = 12 , ⎪− 2 1 + Θ − 2 ΦΘ ⎪ k = 13 ⎪ ρ11 , ⎪ Φρ , k > 13 k −12 ⎩

:‫دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﺒﻌﺾ اﻟﻨﻤﺎذج اﻟﻤﻮﺳﻤﻴﺔ‬

wt = (1 − ΘB s ) at SARIMA(٠,d,٠)(٠,D,١)s ‫ ﻧﻤﻮذج‬-١

k =0 ⎧ 1, ⎪⎪ Θ , k=s ρk = ⎨− 2 1 + Θ ⎪ otherwise ⎪⎩ 0,

(1 − Φ B ) w s

t

= at

SARIMA(٠,d,٠)(١,D,١)s ‫ ﻧﻤﻮذج‬-٢

k =0 ⎧1, ⎪ ks ρk = ⎨Φ , k = s, 2 s,... ⎪ 0, otherwise ⎩

wt = (1 − θ B ) (1 − ΘB s ) at

١٠٨

SARIMA(٠,d,١)(٠,D,١)s ‫ ﻧﻤﻮذج‬-٣

k =0 ⎧ 1, ⎪ θ ⎪− , k =1 2 ⎪ 1+θ ⎪ θΘ , k = s −1 ⎪⎪ 2 2 ρ k = ⎨ (1 + θ )(1 + Θ ) ⎪ Θ ⎪− , k=s 2 ⎪ 1+ Θ ⎪ ρ s−1 , k = s +1 ⎪ otherwise ⎪⎩ 0,

(1 − ΦB ) w = (1 − ΘB ) a s

s

t

t

k =0 ⎧ 1, ⎪ ⎪ ( Θ − Φ )(1 − ΦΘ ) k s−1 ρk = ⎨− Φ , k = s, 2 s,... 2 ⎪ 1 + Θ − 2ΦΘ otherwise ⎪⎩ 0, (1 − ΦB s ) wt = (1 − θ B ) at

SARIMA(٠,d,٠)(١,D,١)s ‫ ﻧﻤﻮذج‬-٤

SARIMA(٠,d,١)(١,D,٠)s ‫ ﻧﻤﻮذج‬-٥

k =0 ⎧ 1, ⎪ θ ⎪− , k =1 2 ⎪ 1+θ ⎪ 0, k = 2,..., s − 2 ⎪ ρk = ⎨ θ Φ − , k = s −1 ⎪ 1+θ 2 ⎪ k=s ⎪ Φ, ⎪ ρ s−1 , k = s +1 ⎪ k > s +1 ⎪⎩ Φ ρ k − s ,

wt = (1 − θ1B − θ 2 B 2 )(1 − ΘB12 ) at

١٠٩

SARIMA(٠,d,٢)(٠,D,١)s ‫ ﻧﻤﻮذج‬-٦

‫‪k =0‬‬ ‫‪k =1‬‬ ‫‪k =2‬‬ ‫‪k = s−2‬‬ ‫‪k = s −1‬‬ ‫‪k=s‬‬ ‫‪k = s +1‬‬ ‫‪k = s+2‬‬ ‫‪otherwise‬‬

‫‪⎧ 1,‬‬ ‫⎪‬ ‫‪⎪ − θ1 (1 − θ 2 ) ,‬‬ ‫‪⎪ 1 + θ12 + θ 22‬‬ ‫⎪‬ ‫‪θ2‬‬ ‫‪⎪−‬‬ ‫‪,‬‬ ‫‪⎪ 1 + θ12 + θ 22‬‬ ‫⎪‬ ‫‪θ 2Θ‬‬ ‫⎪‬ ‫‪,‬‬ ‫‪2‬‬ ‫) ‪⎪⎪ (1 + θ1 + θ 22 )(1 + Θ2‬‬ ‫⎨ = ‪ρk‬‬ ‫) ‪θ1Θ (1 − θ 2‬‬ ‫⎪‬ ‫‪,‬‬ ‫) ‪⎪ (1 + θ 2 + θ 2 )(1 + Θ2‬‬ ‫‪1‬‬ ‫‪2‬‬ ‫⎪‬ ‫⎪‬ ‫‪Θ‬‬ ‫‪,‬‬ ‫‪⎪−‬‬ ‫‪2‬‬ ‫‪+‬‬ ‫‪Θ‬‬ ‫‪1‬‬ ‫⎪‬ ‫‪⎪ ρ s −1 ,‬‬ ‫‪⎪ ρ ,‬‬ ‫‪⎪ s −2‬‬ ‫‪⎪⎩ 0,‬‬

‫ﺳﻮف ﻧﺴﺘﻌﺮض ﺑﻌﺾ اﻟﺮﺳﻮﻣﺎت ﻟﺪاﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻟﻠﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ ﻷﻋﻄﺎء‬ ‫ﻓﻜﺮة ﻋﻦ أﺷﻜﺎﻟﻬﺎ‪.‬‬ ‫اﻷﺷﻜﺎل اﻟﺘﺎﻟﻴﺔ ﻟﻨﻤﻮذج ‪: SARIMA(٠,d,١)(١,D,٠)١٢‬‬ ‫ﺷﻜﻞ )‪(١‬‬

‫‪Φ = 0.6, θ = 0.5‬‬ ‫‪A ( 0 ,d ,1 ) ( 1 ,D ,0 ) 1 2‬‬

‫‪o f S A R I M‬‬

‫‪A C F‬‬ ‫‪0 .5‬‬

‫‪C1‬‬

‫‪0 .0‬‬

‫‪-0 .5‬‬

‫‪5 0‬‬

‫‪4 0‬‬

‫‪2 0‬‬

‫‪3 0‬‬

‫‪0‬‬

‫‪1 0‬‬

‫‪L a g‬‬

‫ﺷﻜﻞ )‪( ٢‬‬

‫‪Φ = 0.6, θ = −0.5‬‬ ‫‪A ( 0 ,d ,1 ) ( 1 ,D ,0 ) 1 2‬‬

‫‪o f S A R I M‬‬

‫‪A C F‬‬ ‫‪0 .7‬‬ ‫‪0 .6‬‬ ‫‪0 .5‬‬

‫‪0 .3‬‬ ‫‪0 .2‬‬ ‫‪0 .1‬‬ ‫‪0 .0‬‬

‫‪5 0‬‬

‫‪4 0‬‬

‫‪2 0‬‬

‫‪3 0‬‬

‫‪L a g‬‬

‫‪١١٠‬‬

‫‪1 0‬‬

‫‪0‬‬

‫‪C1‬‬

‫‪0 .4‬‬

‫ﺷﻜﻞ )‪(٣‬‬

‫‪Φ = −0.6, θ = 0.5‬‬ ‫‪A C F o f S A R IM A (0 ,d ,1 )(1 ,D ,0 )1 2‬‬ ‫‪0 .5‬‬

‫‪C1‬‬

‫‪0 .0‬‬

‫‪-0 .5‬‬

‫‪50‬‬

‫‪40‬‬

‫‪20‬‬

‫‪30‬‬

‫‪10‬‬

‫‪0‬‬

‫‪Lag‬‬

‫ﺷﻜﻞ )‪(٤‬‬

‫‪Φ = −0.6, θ = −0.5‬‬ ‫‪A C F o f S A R IM A (0 ,d ,1 )(1 ,D ,0 )1 2‬‬

‫‪0 .5‬‬

‫‪C1‬‬

‫‪0 .0‬‬

‫‪-0 .5‬‬

‫‪5 0‬‬

‫‪4 0‬‬

‫‪2 0‬‬

‫‪3 0‬‬

‫‪1 0‬‬

‫‪0‬‬

‫‪L ag‬‬

‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻨﻤﻮذج اﻟﻤﻮﺳﻤﻲ اﻟﺘﻀﺎﻋﻔﻲ‪:‬‬ ‫ﻣﻦ اﻟﺼﻌﻮﺑﺔ إﺷﺘﻘﺎق وﺗﻔﺴﻴﺮ أﻧﻤﺎط داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻨﻤﺎذج اﻟﻤﻮﺳﻤﻴﺔ اﻟﺘﻀﺎﻋﻔﻴﺔ‬ ‫وﻟﻜﻨﻬﺎ وﺑﺸﻜﻞ ﻋﺎم ﻓﺈن أﺟﺰاء اﻟﻨﻤﻮذج اﻟﻤﻮﺳﻤﻴﺔ وﻏﻴﺮ اﻟﻤﻮﺳﻤﻴﺔ واﻟﺘﻲ ﺗﻨﻤﺬج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك‬ ‫ﺗﻌﻄﻲ ﺗﺨﺎﻣﺪات اﺳﻴﺔ وﺗﺨﺎﻣﺪات ﺟﻴﺒﻴﺔ ﻋﻨﺪ اﻟﺘﺨﻠﻔﺎت اﻟﻤﻮﺳﻤﻴﺔ وﻏﻴﺮاﻟﻤﻮﺳﻤﻴﺔ وﻓﻲ اﻟﻨﻤﺎذج اﻟﺘﻲ‬ ‫ﺗﺤﻮي إﻧﺤﺪار ذاﺗﻲ ﻓﺈن اﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﺠﺰﺋﻴﺔ ﺗﻌﻄﻲ ﻗﻄﻌﺎ ‪. cut off‬‬ ‫اﻷﺷﻜﺎل اﻟﺘﺎﻟﻴﺔ ﻹﻋﻄﺎء ﻓﻜﺮة ﻋﻦ ﺑﻌﺾ دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﺒﻌﺾ اﻟﻨﻤﺎذج ‪:‬‬ ‫‪ -١‬ﺷﻜﻞ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻨﻤﻮذج ‪wt = (1 − ΘB12 ) at‬‬ ‫ا( ‪Θ = 0.6‬‬

‫‪١١١‬‬

A C F o f S A R IM A (0 ,d ,0 )(0 ,D ,1 )1 2 0 .0 -0 .1

C1

-0 .2 -0 .3 -0 .4 -0 .5 -0 .6

0

1 0

2 0

3 0

4 0

5 0

L ag

Θ = −0.6 (‫ب‬ A C F o f S A R IM A (0 ,d ,0 )(0 ,D ,1 )1 2 0 .6 0 .5 0 .4

C1

0 .3 0 .2 0 .1 0 .0 -0 .1 -0 .2 0

1 0

2 0

3 0

4 0

5 0

L ag

(1 − Φ B ) w 12

t

= at ‫ ﺷﻜﻞ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻨﻤﻮذج‬-٢

Φ = 0.6 (‫ا‬ A C F o f S A R IM A (0 ,d ,1 )(0 ,D ,0 )1 2 0 .6 0 .5

C1

0 .4 0 .3 0 .2 0 .1 0 .0

0

1 0

2 0

3 0

4 0

5 0

L a g

Φ = −0.6 (‫ب‬

١١٢

‫‪A C F o f S A R IM A (0 ,d ,1 )(0 ,D ,0 )1 2‬‬ ‫‪0 .0‬‬ ‫‪-0 .1‬‬ ‫‪-0 .2‬‬

‫‪C1‬‬

‫‪-0 .3‬‬ ‫‪-0 .4‬‬ ‫‪-0 .5‬‬ ‫‪-0 .6‬‬

‫‪5 0‬‬

‫‪4 0‬‬

‫‪2 0‬‬

‫‪3 0‬‬

‫‪0‬‬

‫‪1 0‬‬

‫‪L ag‬‬

‫ﺃﻤﺜﻠﺔ‪ :‬ﻟﻠﻤﺘﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺍﻟﻤﻭﺴﻤﻴﺔ )ﻓﻲ ﺠﻤﻴﻊ ﺍﻷﻤﺜﻠﺔ ﺍﻟﺘﺎﻟﻴﺔ ﺇﻗﺭﺃ ﺴﻁﺭﺍ ﺒﺴﻁﺭ(‬ ‫‪54.9‬‬ ‫‪58.2‬‬ ‫‪68.2‬‬ ‫‪54.9‬‬ ‫‪53.2‬‬ ‫‪57.3‬‬ ‫‪70.4‬‬ ‫‪53.4‬‬ ‫‪51.7‬‬ ‫‪57.1‬‬ ‫‪75.5‬‬

‫‪54.9‬‬ ‫‪54.3‬‬ ‫‪70.1‬‬ ‫‪57.5‬‬ ‫‪52.8‬‬ ‫‪52.7‬‬ ‫‪69.9‬‬ ‫‪55.3‬‬ ‫‪51.5‬‬ ‫‪53.9‬‬ ‫‪73.3‬‬

‫‪55.3‬‬ ‫‪53.4‬‬ ‫‪67.5‬‬ ‫‪61.2‬‬ ‫‪52.8‬‬ ‫‪51.6‬‬ ‫‪61.0‬‬ ‫‪58.2‬‬ ‫‪52.3‬‬ ‫‪53.5‬‬ ‫‪68.1‬‬ ‫‪62.2‬‬

‫‪56.9‬‬ ‫‪53.0‬‬ ‫‪58.0‬‬ ‫‪69.3‬‬ ‫‪54.5‬‬ ‫‪52.4‬‬ ‫‪52.7‬‬ ‫‪66.9‬‬ ‫‪53.6‬‬ ‫‪53.5‬‬ ‫‪58.1‬‬ ‫‪74.8‬‬

‫‪57.4‬‬ ‫‪52.8‬‬ ‫‪54.2‬‬ ‫‪69.8‬‬ ‫‪56.4‬‬ ‫‪52.1‬‬ ‫‪53.9‬‬ ‫‪70.7‬‬ ‫‪55.3‬‬ ‫‪53.1‬‬ ‫‪54.8‬‬ ‫‪76.4‬‬

‫‪61.5‬‬ ‫‪53.3‬‬ ‫‪53.2‬‬ ‫‪66.1‬‬ ‫‪59.3‬‬ ‫‪52.6‬‬ ‫‪53.3‬‬ ‫‪65.3‬‬ ‫‪58.5‬‬ ‫‪53.3‬‬ ‫‪54.3‬‬ ‫‪70.8‬‬

‫‪72.7‬‬ ‫‪54.4‬‬ ‫‪53.0‬‬ ‫‪56.1‬‬ ‫‪68.7‬‬ ‫‪53.9‬‬ ‫‪53.5‬‬ ‫‪56.5‬‬ ‫‪69.3‬‬ ‫‪53.9‬‬ ‫‪54.6‬‬ ‫‪60.6‬‬

‫‪72.2‬‬ ‫‪56.0‬‬ ‫‪53.0‬‬ ‫‪53.6‬‬ ‫‪70.0‬‬ ‫‪56.4‬‬ ‫‪53.4‬‬ ‫‪53.4‬‬ ‫‪69.6‬‬ ‫‪55.6‬‬ ‫‪54.2‬‬ ‫‪56.4‬‬

‫‪71.5‬‬ ‫‪60.0‬‬ ‫‪53.4‬‬ ‫‪53.0‬‬ ‫‪67.9‬‬ ‫‪61.7‬‬ ‫‪53.6‬‬ ‫‪52.5‬‬ ‫‪64.2‬‬ ‫‪60.1‬‬ ‫‪54.8‬‬ ‫‪55.6‬‬

‫‪59.1‬‬ ‫‪71.0‬‬ ‫‪54.6‬‬ ‫‪52.8‬‬ ‫‪58.7‬‬ ‫‪68.3‬‬ ‫‪55.1‬‬ ‫‪53.2‬‬ ‫‪55.5‬‬ ‫‪68.9‬‬ ‫‪55.8‬‬ ‫‪55.0‬‬

‫‪57.2‬‬ ‫‪70.6‬‬ ‫‪55.6‬‬ ‫‪52.6‬‬ ‫‪55.4‬‬ ‫‪67.9‬‬ ‫‪57.3‬‬ ‫‪53.0‬‬ ‫‪53.3‬‬ ‫‪68.8‬‬ ‫‪57.9‬‬ ‫‪54.7‬‬

‫‪56.3‬‬ ‫‪68.2‬‬ ‫‪59.4‬‬ ‫‪52.9‬‬ ‫‪52.9‬‬ ‫‪65.3‬‬ ‫‪61.9‬‬ ‫‪53.5‬‬ ‫‪52.4‬‬ ‫‪63.6‬‬ ‫‪62.6‬‬ ‫‪55.8‬‬

‫‪55.8‬‬ ‫‪57.7‬‬ ‫‪69.8‬‬ ‫‪54.0‬‬ ‫‪53.0‬‬ ‫‪58.2‬‬ ‫‪69.9‬‬ ‫‪54.3‬‬ ‫‪51.5‬‬ ‫‪57.1‬‬ ‫‪70.3‬‬ ‫‪57.7‬‬

‫‪55.7‬‬ ‫‪54.6‬‬ ‫‪71.0‬‬ ‫‪54.9‬‬ ‫‪52.7‬‬ ‫‪55.8‬‬ ‫‪71.5‬‬ ‫‪56.3‬‬ ‫‪51.5‬‬ ‫‪52.2‬‬ ‫‪69.4‬‬ ‫‪60.5‬‬

‫ﺷﻜﻞ اﻟﻤﺘﺴﻠﺴﻠﺔ هﻮ‬

‫‪7 0‬‬

‫)‪z(t‬‬

‫‪6 0‬‬

‫‪5 0‬‬

‫‪1 5 0‬‬

‫‪1 0 0‬‬

‫داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ‬

‫‪١١٣‬‬

‫‪5 0‬‬

‫‪In d e x‬‬

‫)‪z(t‬‬ ‫‪56.3‬‬ ‫‪54.9‬‬ ‫‪67.4‬‬ ‫‪56.6‬‬ ‫‪53.4‬‬ ‫‪55.3‬‬ ‫‪65.1‬‬ ‫‪59.4‬‬ ‫‪52.1‬‬ ‫‪51.5‬‬ ‫‪64.7‬‬ ‫‪66.4‬‬

Autocorrelation

A u t o c o r r e la t io n F u n c t io n f o r z ( t ) 1 .0 0 .8 0 .6 0 .4 0 .2 0 .0 -0 .2 -0 .4 -0 .6 -0 .8 -1 .0

2 Lag 1 2 3 4 5 6 7 8 9 10 11 12

C o rr

12

T

LBQ

Lag

C o rr

0 .7 7 1 0 .3 1 2 .9 3 0 .3 3 -0 .1 1 -0 .9 9 -0 .3 8 -3 .2 8 -0 .5 2 -4 .2 2 -0 .5 6 -4 .1 6 -0 .5 2 -3 .5 0 -0 .3 8 -2 .3 9 -0 .1 2 -0 .7 1 1 .7 6 0 .2 9 4 .1 5 0 .6 9 4 .8 6 0 .8 8

1 0 8 .1 0 1 2 7 .3 4 1 2 9 .7 4 1 5 6 .8 3 2 0 7 .2 4 2 6 6 .5 5 3 1 6 .8 0 3 4 3 .7 1 3 4 6 .2 5 3 6 2 .0 8 4 5 3 .5 8 6 0 3 .8 7

13 14 15 16 17 18 19 20 21 22 23 24

0 .6 8 0 .2 8 -0 .1 2 -0 .3 6 -0 .4 9 -0 .5 3 -0 .4 9 -0 .3 7 -0 .1 4 0 .2 3 0 .6 1 0 .7 9

22 T

32

42

LBQ

Lag

C o rr

T

LBQ

Lag

C o rr

T

LBQ

3 .3 4 6 9 4 .2 8 1 .3 0 7 0 9 .8 2 -0 .5 3 7 1 2 .4 3 -1 .6 5 7 3 8 .3 0 -2 .2 1 7 8 6 .2 8 -2 .3 3 8 4 3 .0 4 -2 .1 0 8 9 2 .3 6 -1 .5 4 9 2 0 .3 9 -0 .5 6 9 2 4 .2 4 0 .9 5 9 3 5 .3 7 2 .4 6 1 0 1 1 .3 1 3 .1 0 1 1 4 0 .9 0

25 26 27 28 29 30 31 32 33 34 35 36

0 .6 1 0 .2 5 -0 .1 1 -0 .3 4 -0 .4 7 -0 .5 1 -0 .4 8 -0 .3 6 -0 .1 4 0 .1 9 0 .5 4 0 .7 2

2 .3 0 0 .9 1 -0 .4 2 -1 .2 4 -1 .6 7 -1 .8 0 -1 .6 5 -1 .2 3 -0 .4 9 0 .6 6 1 .8 2 2 .3 7

1 2 2 0 .0 7 1 2 3 3 .1 8 1 2 3 5 .9 8 1 2 6 1 .2 9 1 3 0 8 .3 2 1 3 6 4 .7 2 1 4 1 4 .0 8 1 4 4 2 .7 1 1 4 4 7 .3 4 1 4 5 5 .7 9 1 5 2 0 .8 9 1 6 3 6 .4 9

37 38 39 40 41 42 43 44

0 .5 7 0 .2 3 -0 .1 1 -0 .3 3 -0 .4 4 -0 .4 8 -0 .4 5 -0 .3 4

1 .8 2 0 .7 3 -0 .3 4 -1 .0 2 -1 .3 8 -1 .4 8 -1 .3 7 -1 .0 4

1 7 0 9 .3 6 1 7 2 1 .7 0 1 7 2 4 .4 3 1 7 4 9 .0 6 1 7 9 5 .0 0 1 8 4 9 .4 1 1 8 9 7 .0 5 1 9 2 5 .3 0

‫واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬ Partial Autocorrelation

P a r t ia l A u t o c o r r e la t io n F u n c t io n f o r z ( t ) 1 .0 0 .8 0 .6 0 .4 0 .2 0 .0 - 0 .2 - 0 .4 - 0 .6 - 0 .8 - 1 .0

2

12

22

32

42

Lag

P AC

T

Lag

P AC

T

Lag

P AC

T

Lag

P AC

T

1 2 3 4 5 6 7 8 9 10 11 12

0 .7 7 -0 .6 8 -0 .0 6 0 .0 1 -0 .4 2 -0 .1 8 -0 .1 1 -0 .2 2 0 .1 8 0 .5 7 0 .2 6 0 .1 6

1 0 .3 1 - 9 .0 1 - 0 .8 2 0 .0 7 - 5 .5 5 - 2 .4 5 - 1 .4 8 - 2 .9 6 2 .3 9 7 .6 5 3 .5 0 2 .1 6

13 14 15 16 17 18 19 20 21 22 23 24

- 0 .4 2 0 .2 9 0 .0 1 - 0 .0 7 0 .0 7 - 0 .0 1 - 0 .0 9 - 0 .0 5 - 0 .0 8 0 .1 0 - 0 .0 3 0 .0 3

- 5 .6 2 3 .8 9 0 .1 6 - 0 .9 2 0 .9 5 - 0 .1 1 - 1 .1 5 - 0 .7 2 - 1 .1 1 1 .3 8 - 0 .3 5 0 .3 7

25 26 27 28 29 30 31 32 33 34 35 36

- 0 .1 3 - 0 .0 0 0 .0 4 - 0 .1 0 - 0 .0 3 0 .0 1 0 .0 1 - 0 .0 4 - 0 .0 4 0 .0 1 - 0 .0 2 0 .0 7

-1 .7 5 -0 .0 5 0 .5 0 -1 .3 0 -0 .3 4 0 .1 9 0 .0 7 -0 .5 5 -0 .5 2 0 .1 5 -0 .3 2 0 .9 0

37 38 39 40 41 42 43 44

-0 .0 6 -0 .0 5 0 .0 1 0 .0 0 -0 .0 5 0 .0 5 -0 .0 5 -0 .0 3

-0 .7 4 -0 .6 2 0 .1 7 0 .0 0 -0 .6 3 0 .6 6 -0 .6 4 -0 .4 2

.‫ﻧﻼﺣﻆ اﻷﻧﻤﺎط اﻟﻤﻮﺳﻤﻴﺔ واﺿﺤﺔ ﻓﻲ اﻷﺷﻜﺎل اﻟﺴﺎﺑﻘﺔ‬ ‫ﻤﺜﺎل ﺁﺨﺭ‬ z(t) 589 673 678 621 713 796 801 747 826 898 908 827

561 742 639 602 667 858 764 711 799 957 867 797

640 716 604 635 762 826 725 751 890 924 815 843

656 660 611 677 784 783 723 804 900 881 812

727 617 594 635 837 740 690 756 961 837 773

697 583 634 736 817 701 734 860 935 784 813

640 587 658 755 767 706 750 878 894 791 834

599 565 622 811 722 677 707 942 855 760 782

568 598 709 798 681 711 807 913 809 802 892

577 628 722 735 687 734 824 869 810 828 903

553 618 782 697 660 690 886 834 766 778 966

582 688 756 661 698 785 859 790 805 889 937

600 705 702 667 717 805 819 800 821 902 896

566 770 653 645 696 871 783 763 773 969 858

653 736 615 688 775 845 740 800 883 947 817

‫ﺷﻜﻞ اﻟﻤﺘﺴﻠﺴﻠﺔ‬

١١٤

1000

z(t)

900

800

700

600

50

In d e x

100

150

‫اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ‬ Autocorrelation

A u t o c o r r e la t io n F u n c t io n f o r z ( t ) 1 0 0 0 0 0 -0 -0 -0 -0 -1

.0 .8 .6 .4 .2 .0 .2 .4 .6 .8 .0

2 Lag 1 2 3 4 5 6 7 8 9 10 11 12

C o rr 0 0 0 0 0 0 0 0 0 0 0 0

.8 .7 .6 .4 .4 .3 .4 .4 .5 .6 .7 .8

T

12 LB Q

9 1 1 .5 6 1 3 5 .9 4 6 .2 7 2 4 0 .1 3 8 4 .1 2 3 0 6 .7 2 2 2 .9 5 3 4 7 .9 7 9 2 .4 7 3 8 0 .0 9 3 2 .1 0 4 0 5 .0 2 8 2 .2 5 4 3 5 .5 4 1 2 .4 0 4 7 2 .3 7 5 2 .8 7 5 2 9 .0 7 6 3 .3 4 6 1 4 .2 7 9 3 .5 2 7 2 1 .7 2 7 3 .6 1 8 5 2 .4 1 4

L ag 1 1 1 1 1 1 1 2 2 2 2 2

3 4 5 6 7 8 9 0 1 2 3 4

22

C o rr 0 0 0 0 0 0 0 0 0 0 0 0

.7 .6 .4 .3 .3 .2 .2 .3 .4 .5 .6 .6

4 4 9 6 1 5 9 2 2 3 0 7

T 2 2 1 1 1 0 1 1 1 1 2 2

.9 .4 .7 .3 .0 .9 .0 .1 .4 .8 .0 .2

6 11 91 11 91 01 11 21 41 11 31 11

LB Q 9 0 0 0 1 1 1 1 1 2 3 4

5 3 7 9 1 2 4 6 9 5 2 1

4 0 4 9 7 9 5 5 9 3 5 5

.6 .0 .8 .6 .3 .7 .5 .4 .1 .8 .5 .2

8 9 5 8 9 6 9 2 3 1 1 9

L ag 2 2 2 2 2 3 3 3 3 3 3 3

5 6 7 8 9 0 1 2 3 4 5 6

32

C o rr 0 0 0 0 0 0 0 0 0 0 0 0

.5 .4 .3 .2 .1 .1 .1 .2 .2 .3 .4 .5

8 9 5 4 9 4 7 0 8 8 5 2

T 1 1 1 0 0 0 0 0 0 1 1 1

.8 .5 .0 .7 .5 .4 .5 .6 .8 .1 .3 .5

61 21 91 31 71 31 11 01 51 51 51 31

LB Q 4 5 5 5 5 5 5 5 6 6 6 7

8 3 5 6 7 7 8 9 1 4 8 4

3 0 6 8 5 9 5 3 0 1 5 2

.1 .9 .4 .2 .4 .5 .5 .6 .3 .5 .2 .9

1 9 5 0 4 5 2 2 5 7 7 5

L ag 3 3 3 4 4 4

7 8 9 0 1 2

42 C o rr 0 0 0 0 0 0

.4 .3 .2 .1 .0 .0

3 5 2 2 6 2

T 1 1 0 0 0 0

.2 .0 .6 .3 .1 .0

71 01 41 31 81 51

LB Q 7 8 8 8 8 8

8 1 2 2 2 2

4 0 1 4 5 5

.1 .6 .6 .6 .5 .6

7 0 8 7 6 2

‫واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬ Partial Autocorrelation

P a r t ia l A u t o c o r r e la t io n F u n c t io n f o r z ( t ) 1 .0 0 .8 0 .6 0 .4 0 .2 0 .0 -0 .2 -0 .4 -0 .6 -0 .8 -1 .0

2

12

22

32

42

L ag

PAC

T

L ag

PAC

T

L ag

PAC

T

L ag

PAC

T

1 2 3 4 5 6 7 8 9 10 11 12

0 .8 9 -0 .0 8 -0 .2 8 0 .0 3 0 .3 5 -0 .0 8 0 .2 8 0 .0 9 0 .4 0 0 .3 0 0 .0 6 0 .2 2

1 1 .5 6 -1 .0 6 -3 .6 5 0 .4 2 4 .5 4 -1 .0 7 3 .6 7 1 .1 9 5 .1 7 3 .9 5 0 .8 1 2 .8 8

13 14 15 16 17 18 19 20 21 22 23 24

-0 .6 3 -0 .0 2 0 .0 7 -0 .0 4 -0 .0 9 -0 .0 4 -0 .0 5 0 .0 3 0 .0 4 0 .0 5 0 .0 5 0 .0 5

-8 .1 9 -0 .2 1 0 .9 5 -0 .5 2 -1 .1 2 -0 .4 9 -0 .6 0 0 .3 8 0 .4 6 0 .6 7 0 .6 0 0 .5 9

25 26 27 28 29 30 31 32 33 34 35 36

-0 .1 8 0 .0 8 0 .0 6 -0 .0 3 -0 .0 4 0 .0 0 -0 .0 6 -0 .0 1 -0 .0 1 0 .0 3 0 .0 0 0 .0 1

-2 .3 6 1 .0 6 0 .7 3 -0 .4 4 -0 .4 7 0 .0 2 -0 .7 2 -0 .1 1 -0 .1 8 0 .3 8 0 .0 3 0 .0 9

37 38 39 40 41 42

-0 .1 1 -0 .0 2 0 .0 4 -0 .0 3 -0 .0 8 0 .0 1

-1 .3 7 -0 .2 2 0 .5 1 -0 .4 2 -1 .0 6 0 .0 8

١١٥

‫ﻣﺜﺎل ﺁﺧﺮ‬ z(t) 302 107 055 237 079 035 256

262 056 048 247 045 056 250

218 049 115 215 040 097 198

175 047 185 182 038 210 136

100 047 276 080 041 260 073

077 071 220 046 069 257 039

043 151 181 065 152 210 032

047 244 151 040 232 125 030

049 280 083 044 282 080 031

069 230 055 063 255 042 045

152 185 049 085 161 035

205 148 042 185 107 031

246 098 046 247 053 032

294 061 074 231 040 050

242 046 103 167 039 092

181 045 200 117 034 189

‫وﻟﻬﺎ اﻟﺸﻜﻞ‬

3 0 0

z(t)

2 0 0

1 0 0

0

1 0

In d e x

2 0

3 0

4 0

5 0

6 0

7 0

8 0

9 0

1 0 0

‫وداﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ‬ Autocorrelation

A u t o c o r r e la t io n F u n c t io n f o r z ( t ) 1 0 0 0 0 0 -0 -0 -0 -0 -1

.0 .8 .6 .4 .2 .0 .2 .4 .6 .8 .0

5

1 2 3 4 5 6 7

C o rr 0 0 -0 -0 -0 -0 -0

.8 .4 .0 .4 .6 .7 .6

1 3 3 3 9 8 8

T 8 2 -0 -2 -4 -3 -3

.3 .8 .1 .7 .0 .9 .0

8 9 7 0 3 9 9

L B Q

1 1 2 2

7 9 9 1 6 3 8

2 2 2 3 6 5 9

.2 .3 .4 .1 .3 .1 .3

L a g

2 2 0 3 7 7 3

1 1 1 1 1

C o rr

T

8 -0 .4 3 -1 .7 8 3 1 9 -0 .0 4 -0 .1 8 3 1 0 0 .3 8 1 .5 4 3 2 1 0 .7 1 2 .8 2 3 8 2 0 .8 4 3 .1 0 4 7 3 0 .7 1 2 .4 2 5 3 4 0 .3 7 1 .2 0 5 5

L B Q 0 0 8 9 6 8 6

.6 .8 .2 .7 .2 .8 .1

L a g

6 8 2 2 2 4 0

1 1 1 1 1 2 2

5 6 7 8 9 0 1

2 5

C o rr -0 -0 -0 -0 -0 -0 -0

.0 .3 .6 .6 .5 .3 .0

3 8 0 7 9 7 4

T -0 -1 -1 -2 -1 -1 -0

.0 .1 .8 .0 .7 .0 .1

8 9 8 3 2 5 0

L B Q 5 5 6 6 7 7 7

5 7 2 7 2 4 4

6 4 0 8 4 2 2

.1 .0 .2 .3 .1 .2 .3

8 7 2 6 7 1 8

L a g 2 2 2 2 2

2 3 4 5 6

C o rr 0 0 0 0 0

.3 .6 .7 .5 .3

3 1 1 9 1

T 0 1 1 1 0

.9 .7 .9 .5 .8

2 1 3 5 0

L B Q 7 8 8 9 9

5 0 7 2 4

6 8 9 8 2

.9 .9 .4 .4 .4

8 7 2 1 6

‫وداﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬ P a r t ia l A u t o c o r r e la t io n F u n c t io n f o r z ( t ) Partial Autocorrelation

L a g

1 5

1 0 0 0 0 0 -0 -0 -0 -0 -1

.0 .8 .6 .4 .2 .0 .2 .4 .6 .8 .0

5

L a g 1 2 3 4 5 6 7

P A C 0 -0 -0 -0 -0 -0 -0

.8 .7 .2 .2 .1 .3 .1

1 0 9 2 6 2 6

T 8 -7 -2 -2 -1 -3 -1

.3 .1 .9 .2 .6 .2 .6

8 7 8 8 1 9 0

1 5

L a g

1 1 1 1 1

P A C

8 -0 .0 3 9 0 .1 9 0 0 .2 7 1 0 .1 8 2 0 .0 7 3 -0 .1 0 4 -0 .1 5

-0 1 2 1 0 -0 -1

.3 .9 .7 .8 .7 .9 .5

T

L a g

1 1 6 2 4 9 2

1 1 1 1 1 2 2

١١٦

2 5

P A C

0 .1 5 0 .0 6 0 .1 7 0 .0 8 9 -0 .0 0 -0 .0 0 .0 1

9 8 1 4 2 3 4

1 0 1 0 -0 -0 0

.9 .8 .1 .3 .2 .2 .4

T

L a g

3 4 0 9 2 9 6

2 2 2 2 2

P A C

2 0 .0 3 0 .0 4 -0 .0 5 -0 .0 6 0 .1

1 7 7 2 0

T 0 0 -0 -0 1

.1 .7 .7 .2 .0

5 0 2 3 1

‫وآﻞ هﺬﻩ اﻟﻤﺘﻠﺴﻼت ﺗﺒﺪي اﻧﻤﺎط ﻣﻮﺳﻤﻴﺔ واﺿﺤﺔ‪.‬‬

‫إﺷﺘﻘﺎق دوال ﺗﻨﺒﺆ ﻟﺒﻌﺾ ﻧﻤﺎذج اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ اﻟﺘﻀﺎﻋﻔﻴﺔ‪:‬‬ ‫ﺑﻤﺎ ان ﻧﻤﺎذج اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ هﻲ ﺣﺎﻟﺔ ﺧﺎﺻﺔ ﻣﻦ ﻧﻤﺎذج ‪ ARIMA‬ﻓﺈن ﻃﺮق‬ ‫اﻟﺘﻌﺎﻣﻞ ﻣﻌﻬﺎ هﻲ ﻧﻔﺲ اﻟﻄﺮق اﻟﺴﺎﺑﻘﺔ ﻣﻦ ﺣﻴﺚ اﻟﺘﻌﺮف ﻋﻠﻲ ﺷﻜﻞ اﻟﻨﻤﻮذج وﺗﻘﺪﻳﺮ ﻣﻌﺎﻟﻢ اﻟﻨﻤﻮذج‬ ‫واﻹﺧﺘﺒﺎرات اﻟﺘﻔﺤﺼﻴﺔ وﻣﻦ ﺛﻢ اﻟﺘﻨﺒﺆ‪ .‬ﺟﻤﻴﻊ اﻟﻄﺮق واﻟﻤﻌﺎدﻻت اﻟﺘﻲ درﺳﻨﺎهﺎ ﺳﺎﺑﻘﺎ ﻟﻠﻨﻤﺎذج ﻏﻴﺮ‬ ‫اﻟﻤﻮﺳﻤﻴﺔ ﺗﻨﻄﺒﻖ هﻨﺎ‪ .‬ﺳﻮف ﻧﺸﺘﻖ دوال اﻟﺘﻨﺒﺆ ﻟﺒﻌﺾ اﻟﻨﻤﺎذج ﻟﻠﺘﻮﺿﻴﺢ ﻓﻘﻂ‪.‬‬ ‫‪ -١‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻠﻨﻤﻮذج ‪: SARIMA(٠،٠،٠)(٠،١،١)١٢‬‬ ‫وﻳﻜﺘﺐ اﻟﻨﻤﻮذج ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫‪(1 − B ) z = (1 − ΘB ) a‬‬ ‫‪12‬‬

‫‪t‬‬

‫‪12‬‬

‫‪t‬‬

‫ﻣﻦ اﻟﻤﻌﺎدﻟﺔ اﻟﻔﺮوﻗﻴﺔ‬ ‫‪zn + l = zn +l−12 + an + l − Θan + l−12‬‬

‫ﻳﻤﻜﻦ اﻟﺤﺼﻮل ﻋﻠﻰ اﻟﺘﻨﺒﺆات آﺎﻟﺘﺎﻟﻲ‬

‫‪zn (1) = zn −11 − Θan −11‬‬ ‫‪zn ( 2 ) = zn −10 − Θan −10‬‬

‫‪M‬‬

‫‪zn (12 ) = zn − Θan‬‬ ‫‪zn ( l ) = zn ( l − 12 ) , l ≥ 12‬‬

‫أو‬ ‫‪l = 1, 2,...,12‬‬

‫‪l > 12‬‬

‫واﺿﺢ أن‬

‫‪⎧ zn + l−12 − Θan + l−12 ,‬‬ ‫⎨ = ) ‪zn ( l‬‬ ‫‪⎩ zn ( l − 12 ) ,‬‬ ‫‪zn (1) = zn (13) = zn ( 25) = L‬‬

‫‪zn ( 2 ) = zn (14 ) = zn ( 26 ) = L‬‬

‫‪M‬‬

‫‪zn (12 ) = zn ( 24 ) = zn ( 36 ) = L‬‬

‫ﺗﺒﺎﻳﻦ أﺧﻄﺎء اﻟﺘﻨﺒﺆ‬

‫)‬

‫‪2‬‬ ‫‪l −1‬‬

‫‪+ L +ψ‬‬

‫‪2‬‬ ‫‪1‬‬

‫‪(1 + ψ‬‬

‫‪2‬‬

‫‪V ⎣⎡ en ( l )⎦⎤ = σ‬‬

‫وداﻟﺔ اﻷوزان ﺗﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ )ﺑﺮهﻦ ذﻟﻚ(‬ ‫‪j = 12, 24,...‬‬ ‫‪otherwise‬‬

‫‪⎧1 − Θ,‬‬ ‫‪⎩ 0,‬‬

‫⎨= ‪ψj‬‬

‫وﺑﺘﻌﻮﻳﺾ اﻷوزان ﻓﻲ ﺻﻴﻐﺔ ﺗﺒﺎﻳﻦ أﺧﻄﺎء اﻟﺘﻨﺒﺆ ﻧﺠﺪ‬ ‫⎥‪⎧ ⎢ l − 1‬‬ ‫⎫‪2‬‬ ‫⎢ ‪V ⎡⎣ en ( l )⎤⎦ = σ 2 ⎨1 +‬‬ ‫⎬ ) ‪(1 − Θ‬‬ ‫⎥‬ ‫⎦ ‪⎩ ⎣ 12‬‬ ‫⎭‬

‫ﺣﻴﺚ ⎦⎥ ‪ ⎢⎣ x‬ﺗﻌﻨﻲ اﻟﺠﺰء اﻟﺼﺤﻴﺢ ﻣﻦ ‪. x‬‬ ‫‪ -٢‬داﻟﺔ اﻟﺘﻨﺒﺆ ﻟﻠﻨﻤﻮذج ‪: SARIMA(٠،١،١)(٠،١،١)١٢‬‬ ‫وﻳﻜﺘﺐ اﻟﻨﻤﻮذج ﻋﻠﻰ اﻟﺸﻜﻞ‬

‫‪(1 − B ) (1 − B12 ) zt = (1 − θ B ) (1 − ΘB12 ) at‬‬

‫‪١١٧‬‬

‫ﻣﻦ اﻟﻤﻌﺎدﻟﺔ اﻟﻔﺮوﻗﻴﺔ‬

‫‪zn + l = zn +l−1 + zn + l−12 − zn + l−13 + an + l − θ an +l−1 − Θan + l−12 + θ Θan+ l−13‬‬

‫ﻳﻤﻜﻦ اﻟﺤﺼﻮل ﻋﻠﻰ اﻟﺘﻨﺒﺆات آﺎﻟﺘﺎﻟﻲ‬

‫‪zn (1) = zn + zn −11 − zn −13 − θ an − Θan −11 + θ Θan −12‬‬ ‫‪zn ( 2 ) = zn (1) + zn −10 − zn −11 − Θan −10 + θ Θan −11‬‬ ‫‪M‬‬

‫‪zn (12 ) = zn (11) + zn − zn −1 − Θan + θ Θan −1‬‬ ‫‪zn (13) = zn (12 ) + zn (1) − zn + θ Θan‬‬ ‫)‪zn ( l ) = zn ( l − 1) + zn ( l − 12 ) − zn ( l − 13‬‬

‫وهﻜﺬا ﺑﻘﻴﻢ أوﻟﻴﺔ‬

‫‪zn (1) = zn + zn −11 − zn −13 − θ an − Θan −11 + θ Θan −12‬‬ ‫‪zn ( 2 ) = zn (1) + zn −10 − zn −11 − Θan −10 + θ Θan −11‬‬

‫‪M‬‬

‫‪zn (12 ) = zn (11) + zn − zn −1 − Θan + θ Θan −1‬‬ ‫‪zn (13) = zn (12 ) + zn (1) − zn + θ Θan‬‬

‫وﻋﻼﻗﺔ ﺗﻜﺮارﻳﺔ‬

‫‪zn ( l ) = zn ( l − 1) + zn ( l − 12 ) − zn ( l − 13) , l > 13‬‬

‫ﻳﻤﻜﻦ ﺗﻮﻟﻴﺪ اﻟﻌﺪد اﻟﻤﻄﻠﻮب ﻣﻦ اﻟﺘﻨﺒﺆات‪.‬‬ ‫أﻣﺜﻠﺔ وﺣﺎﻻت دراﺳﺔ ﻟﺒﻌﺾ اﻟﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ‪:‬‬ ‫ﻣﺜﺎل )‪ : (١‬ﺳﻮف ﻧﺤﺎول إﻳﺠﺎد ﻧﻤﻮذج ﻣﻦ ﻋﺎﺋﻠﺔ ‪ SARIMA‬ﻳﻨﻄﺒﻖ ﻋﻠﻰ اﻟﻤﺸﺎهﺪات اﻟﺘﺎﻟﻴﺔ‪:‬‬ ‫‪653‬‬ ‫‪736‬‬ ‫‪615‬‬ ‫‪688‬‬ ‫‪775‬‬ ‫‪845‬‬ ‫‪740‬‬ ‫‪800‬‬ ‫‪883‬‬ ‫‪947‬‬ ‫‪817‬‬

‫‪566‬‬ ‫‪770‬‬ ‫‪653‬‬ ‫‪645‬‬ ‫‪696‬‬ ‫‪871‬‬ ‫‪783‬‬ ‫‪763‬‬ ‫‪773‬‬ ‫‪969‬‬ ‫‪858‬‬

‫‪600‬‬ ‫‪705‬‬ ‫‪702‬‬ ‫‪667‬‬ ‫‪717‬‬ ‫‪805‬‬ ‫‪819‬‬ ‫‪800‬‬ ‫‪821‬‬ ‫‪902‬‬ ‫‪896‬‬

‫‪582‬‬ ‫‪688‬‬ ‫‪756‬‬ ‫‪661‬‬ ‫‪698‬‬ ‫‪785‬‬ ‫‪859‬‬ ‫‪790‬‬ ‫‪805‬‬ ‫‪889‬‬ ‫‪937‬‬

‫‪553‬‬ ‫‪618‬‬ ‫‪782‬‬ ‫‪697‬‬ ‫‪660‬‬ ‫‪690‬‬ ‫‪886‬‬ ‫‪834‬‬ ‫‪766‬‬ ‫‪778‬‬ ‫‪966‬‬

‫‪577‬‬ ‫‪628‬‬ ‫‪722‬‬ ‫‪735‬‬ ‫‪687‬‬ ‫‪734‬‬ ‫‪824‬‬ ‫‪869‬‬ ‫‪810‬‬ ‫‪828‬‬ ‫‪903‬‬

‫‪568‬‬ ‫‪598‬‬ ‫‪709‬‬ ‫‪798‬‬ ‫‪681‬‬ ‫‪711‬‬ ‫‪807‬‬ ‫‪913‬‬ ‫‪809‬‬ ‫‪802‬‬ ‫‪892‬‬

‫‪599‬‬ ‫‪565‬‬ ‫‪622‬‬ ‫‪811‬‬ ‫‪722‬‬ ‫‪677‬‬ ‫‪707‬‬ ‫‪942‬‬ ‫‪855‬‬ ‫‪760‬‬ ‫‪782‬‬

‫واﻟﻤﺨﻄﻂ اﻟﺰﻣﻨﻲ ﻟﻠﻤﺸﺎهﺪات‬

‫‪١١٨‬‬

‫‪640‬‬ ‫‪587‬‬ ‫‪658‬‬ ‫‪755‬‬ ‫‪767‬‬ ‫‪706‬‬ ‫‪750‬‬ ‫‪878‬‬ ‫‪894‬‬ ‫‪791‬‬ ‫‪834‬‬

‫‪697‬‬ ‫‪583‬‬ ‫‪634‬‬ ‫‪736‬‬ ‫‪817‬‬ ‫‪701‬‬ ‫‪734‬‬ ‫‪860‬‬ ‫‪935‬‬ ‫‪784‬‬ ‫‪813‬‬

‫‪727‬‬ ‫‪617‬‬ ‫‪594‬‬ ‫‪635‬‬ ‫‪837‬‬ ‫‪740‬‬ ‫‪690‬‬ ‫‪756‬‬ ‫‪961‬‬ ‫‪837‬‬ ‫‪773‬‬

‫‪656‬‬ ‫‪660‬‬ ‫‪611‬‬ ‫‪677‬‬ ‫‪784‬‬ ‫‪783‬‬ ‫‪723‬‬ ‫‪804‬‬ ‫‪900‬‬ ‫‪881‬‬ ‫‪812‬‬

‫‪640‬‬ ‫‪716‬‬ ‫‪604‬‬ ‫‪635‬‬ ‫‪762‬‬ ‫‪826‬‬ ‫‪725‬‬ ‫‪751‬‬ ‫‪890‬‬ ‫‪924‬‬ ‫‪815‬‬ ‫‪843‬‬

‫‪561‬‬ ‫‪742‬‬ ‫‪639‬‬ ‫‪602‬‬ ‫‪667‬‬ ‫‪858‬‬ ‫‪764‬‬ ‫‪711‬‬ ‫‪799‬‬ ‫‪957‬‬ ‫‪867‬‬ ‫‪797‬‬

‫)‪z(t‬‬ ‫‪589‬‬ ‫‪673‬‬ ‫‪678‬‬ ‫‪621‬‬ ‫‪713‬‬ ‫‪796‬‬ ‫‪801‬‬ ‫‪747‬‬ ‫‪826‬‬ ‫‪898‬‬ ‫‪908‬‬ ‫‪827‬‬

‫‪1000‬‬

‫‪900‬‬

‫)‪z(t‬‬

‫‪800‬‬

‫‪700‬‬

‫‪600‬‬

‫‪150‬‬

‫‪50‬‬

‫‪100‬‬

‫‪In d e x‬‬

‫ﻳﻼﺣﻆ ان اﻟﻤﺘﺴﻠﺴﻠﺔ ﻏﻴﺮ ﻣﺴﺘﻘﺮة ﻓﻲ اﻟﺘﺒﺎﻳﻦ واﻟﻤﺘﻮﺳﻂ ﻟﺬﻟﻚ ﻧﺜﺒﺖ اﻟﺘﺒﺎﻳﻦ أوﻻ ﺑﺘﺤﻮﻳﻞ ﻟﻮﻏﺎرﺛﻤﻲ‬ ‫أي ) ‪ yt = ln ( zt‬وﻧﺮﺳﻢ اﻟﻤﺨﻄﻂ اﻟﺰﻣﻨﻲ ﻟﻬﺎ‬ ‫‪6 .9‬‬ ‫‪6 .8‬‬ ‫‪6 .7‬‬

‫)‪y(t‬‬

‫‪6 .6‬‬ ‫‪6 .5‬‬ ‫‪6 .4‬‬ ‫‪6 .3‬‬

‫‪1 5 0‬‬

‫‪5 0‬‬

‫‪1 0 0‬‬

‫‪In d e x‬‬

‫ﻧﻼﺣﻆ ان اﻟﻤﺘﺴﻠﺴﻠﺔ اﺳﺘﻘﺮت ﻓﻲ اﻟﺘﺒﺎﻳﻦ وﻟﻜﻦ ﻻﺗﺰال ﻏﻴﺮ ﻣﺴﺘﻘﺮة ﻓﻲ اﻟﻤﺘﻮﺳﻂ ﻟﺬﻟﻚ ﻧﺄﺧﺬ اﻟﻔﺮق‬ ‫اﻷول ) ‪ xt = (1 − B ) yt = (1 − B ) ln ( zt‬وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‬ ‫‪0 .1 5‬‬ ‫‪0 .1 0‬‬

‫‪0 .0 0‬‬

‫)‪y(t)-y(t-1‬‬

‫‪0 .0 5‬‬

‫‪-0 .0 5‬‬ ‫‪-0 .1 0‬‬ ‫‪1 5 0‬‬

‫‪1 0 0‬‬

‫‪5 0‬‬

‫‪In d e x‬‬

‫اﻟﻤﺘﺴﻠﺴﻠﺔ اﻵن ﻣﺴﺘﻘﺮة ﻓﻲ آﻞ ﻣﻦ اﻟﺘﺒﺎﻳﻦ واﻟﻤﺘﻮﺳﻂ‪ .‬ﻟﻨﻨﻈﺮ إﻟﻰ دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ‬ ‫اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻬﺎ‬

‫‪١١٩‬‬

Autocorrelation

A u t o c o r r e la t io n F u n c t io n f o r y ( t ) - y ( t 1 .0 0 .8 0 .6 0 .4 0 .2 0 .0 - 0 .2 - 0 .4 - 0 .6 - 0 .8 - 1 .0

10

20

30

L ag

C o rr

T

LBQ

L ag

C o rr

T

LBQ

L ag

C o rr

T

LBQ

1 2 3 4 5 6 7 8 9 10 11 12

0 .0 1 0 .2 5 - 0 .0 8 - 0 .3 7 - 0 .0 6 - 0 .5 0 - 0 .0 4 - 0 .3 5 - 0 .0 5 0 .2 3 0 .0 1 0 .9 0

0 .1 2 3 .2 3 - 1 .0 1 - 4 .4 8 - 0 .6 8 - 5 .4 1 - 0 .3 3 - 3 .3 0 - 0 .4 3 2 .0 1 0 .1 1 7 .7 4

0 .0 2 1 0 .7 0 1 1 .8 8 3 5 .6 0 3 6 .2 8 7 9 .9 7 8 0 .1 9 1 0 2 .3 7 1 0 2 .8 0 1 1 2 .2 9 1 1 2 .3 2 2 6 1 .4 1

13 14 15 16 17 18 19 20 21 22 23 24

0 .0 2 0 .2 3 - 0 .0 7 - 0 .3 4 - 0 .0 6 - 0 .4 6 - 0 .0 3 - 0 .3 2 - 0 .0 5 0 .2 1 0 .0 1 0 .8 2

0 .1 2 1 .5 1 - 0 .4 7 - 2 .1 8 - 0 .3 9 - 2 .8 8 - 0 .2 0 - 1 .9 3 - 0 .2 6 1 .2 0 0 .0 8 4 .7 6

2 6 1 .4 7 2 7 1 .2 9 2 7 2 .3 0 2 9 3 .7 3 2 9 4 .4 5 3 3 4 .6 7 3 3 4 .8 8 3 5 4 .9 6 3 5 5 .3 6 3 6 3 .5 5 3 6 3 .5 8 4 9 7 .3 0

25 26 27 28 29 30 31 32 33 34 35 36

0 .0 2 0 .2 1 - 0 .0 6 - 0 .3 1 - 0 .0 6 - 0 .4 2 - 0 .0 2 - 0 .3 0 - 0 .0 5 0 .1 8 0 .0 0 0 .7 6

0 .0 8 1 .1 0 - 0 .2 9 - 1 .5 7 - 0 .2 9 - 2 .0 9 - 0 .1 2 - 1 .4 7 - 0 .2 4 0 .8 9 0 .0 2 3 .6 3

4 9 7 .3 5 5 0 6 .5 0 5 0 7 .1 6 5 2 6 .4 5 5 2 7 .1 4 5 6 3 .0 5 5 6 3 .1 6 5 8 2 .0 5 5 8 2 .5 8 5 8 9 .8 4 5 8 9 .8 5 7 1 3 .1 4

40

L ag

T

LBQ

3 7 0 .0 1 0 .0 5 3 8 0 .2 0 0 .9 0 3 9 - 0 .0 4 - 0 .1 9 4 0 - 0 .2 7 - 1 .2 1 4 1 - 0 .0 5 - 0 .2 4

C o rr

7 1 3 .1 6 7 2 2 .0 6 7 2 2 .4 5 7 3 8 .8 6 7 3 9 .5 1

Partial Autocorrelation

P a r t ia l A u t o c o r r e la t i o n F u n c t io n f o r y ( t ) - y ( t 1 .0 0 .8 0 .6 0 .4 0 .2 0 .0 -0 .2 -0 .4 -0 .6 -0 .8 -1 .0

10

20

30

40

L ag

PAC

T

L ag

PAC

T

L ag

PAC

T

L ag

PAC

T

1 2 3 4 5 6 7 8 9 10 11 12

0 .0 1 0 .2 5 - 0 .0 9 - 0 .4 6 - 0 .0 2 - 0 .3 5 - 0 .1 4 - 0 .4 8 - 0 .4 1 - 0 .2 3 - 0 .5 8 0 .6 3

0 .1 2 3 .2 3 - 1 .2 0 - 5 .9 7 - 0 .2 2 - 4 .5 2 - 1 .8 2 - 6 .2 6 - 5 .3 6 - 2 .9 9 - 7 .5 5 8 .0 8

13 14 15 16 17 18 19 20 21 22 23 24

-0 .0 4 -0 .3 3 0 .0 0 0 .1 8 0 .0 1 0 .0 8 -0 .0 8 0 .0 8 0 .0 2 -0 .0 4 0 .0 2 0 .0 8

- 0 .4 8 - 4 .2 6 0 .0 6 2 .2 9 0 .0 7 1 .0 6 - 1 .0 4 0 .9 8 0 .2 9 - 0 .5 3 0 .2 9 1 .0 4

25 26 27 28 29 30 31 32 33 34 35 36

- 0 .1 3 - 0 .0 5 0 .0 6 0 .0 0 - 0 .0 7 0 .0 3 0 .0 7 - 0 .0 1 - 0 .1 0 0 .0 6 0 .0 3 0 .0 1

-1 .6 7 -0 .6 1 0 .8 4 0 .0 5 -0 .9 4 0 .3 4 0 .9 3 -0 .0 9 -1 .3 1 0 .8 0 0 .3 9 0 .1 3

37 38 39 40 41

- 0 .0 4 0 .0 2 0 .0 4 0 .0 6 - 0 .0 2

- 0 .5 1 0 .3 0 0 .4 8 0 .8 1 - 0 .2 0

‫ و‬١٢ ‫ﻳﻼﺣﻆ ﻣﻦ داﻟﺔ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ان اﻟﻤﺘﺴﻠﺴﻠﺔ ﻏﻴﺮ ﻣﺴﺘﻘﺮة ﻣﻮﺳﻤﻴﺎ ﻷن ﻗﻴﻤﻬﺎ ﻋﻨﺪ اﻟﺘﺨﻠﻔﺎت‬ wt = (1 − B12 ) (1 − B ) ln ( zt ) ‫ ﺗﺘﺨﺎﻣﺪ ﺑﺒﻂء ﻟﺬﻟﻚ ﻧﺄﺧﺬ اﻟﻔﺮق اﻟﻤﻮﺳﻤﻲ اﻷول‬٣٦ ‫ و‬٢٤ ‫وﻧﺮﺳﻤﻬﺎ ﺑﻌﺪ هﺬا اﻟﺘﻔﺮﻳﻖ‬

y(t)-y(t-1)12

0 .0 5

0 .0 0

-0 .0 5 In d e x

50

100

150

‫ﻧﻮﺟﺪ دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻬﺎ‬

١٢٠

Autocorrelation

A u t o c o r r e la t io n F u n c t io n f o r y ( t ) - y ( t 1 .0 0 .8 0 .6 0 .4 0 .2 0 .0 - 0 .2 - 0 .4 - 0 .6 - 0 .8 - 1 .0

5

15

25

L ag

C o rr

T

LBQ

L ag

C o rr

T

LBQ

L ag

C o rr

T

LBQ

1 2 3 4 5 6 7 8 9 10 11 12

- 0 .2 1 - 0 .0 1 0 .1 0 - 0 .1 3 - 0 .1 0 - 0 .0 2 0 .1 2 0 .0 5 - 0 .0 5 0 .1 3 - 0 .0 1 - 0 .4 4

- 2 .6 5 - 0 .1 3 1 .1 4 - 1 .4 9 - 1 .1 1 - 0 .2 8 1 .3 3 0 .5 4 - 0 .5 8 1 .4 5 - 0 .0 9 - 4 .9 0

7 .1 5 7 .1 7 8 .6 3 1 1 .2 0 1 2 .6 7 1 2 .7 7 1 4 .9 5 1 5 .3 1 1 5 .7 5 1 8 .4 9 1 8 .5 0 5 0 .8 8

13 14 15 16 17 18 19 20 21 22 23 24

0 .1 8 - 0 .0 7 - 0 .0 5 0 .0 3 0 .1 2 - 0 .0 0 - 0 .1 1 0 .0 3 - 0 .0 2 - 0 .0 9 0 .1 1 - 0 .0 4

1 .8 0 - 0 .7 0 - 0 .4 8 0 .2 6 1 .1 3 - 0 .0 2 - 1 .0 8 0 .2 5 - 0 .2 0 - 0 .8 0 0 .9 9 - 0 .3 8

5 6 .6 4 5 7 .5 4 5 7 .9 7 5 8 .1 0 6 0 .5 8 6 0 .5 8 6 2 .9 1 6 3 .0 4 6 3 .1 2 6 4 .4 5 6 6 .5 2 6 6 .8 2

25 26 27 28 29 30 31 32 33 34 35 36

0 .0 7 - 0 .0 0 - 0 .0 6 0 .0 3 - 0 .1 0 0 .0 1 0 .0 4 0 .0 0 0 .0 2 0 .0 0 0 .0 9 - 0 .0 6

0 .6 2 - 0 .0 2 - 0 .5 7 0 .3 0 - 0 .9 0 0 .0 8 0 .3 8 0 .0 4 0 .1 9 0 .0 3 0 .8 0 - 0 .5 1

6 7 .6 7 6 7 .6 7 6 8 .3 8 6 8 .5 8 7 0 .4 0 7 0 .4 1 7 0 .7 4 7 0 .7 4 7 0 .8 3 7 0 .8 3 7 2 .3 6 7 2 .9 9

35 L ag

T

LBQ

3 7 - 0 .0 6 - 0 .5 4 3 8 0 .0 3 0 .2 6

C o rr

7 3 .7 1 7 3 .8 8

Partial Autocorrelation

P a r t ia l A u t o c o r r e la t io n F u n c t io n f o r y ( t ) - y ( t 1 .0 0 .8 0 .6 0 .4 0 .2 0 .0 -0 .2 -0 .4 -0 .6 -0 .8 -1 .0

5

15

25

35

L ag

PAC

T

L ag

PAC

T

L ag

PAC

T

L ag

PAC

T

1 2 3 4 5 6 7 8 9 10 11 12

- 0 .2 1 - 0 .0 6 0 .0 8 - 0 .0 9 - 0 .1 5 - 0 .1 0 0 .1 1 0 .1 1 - 0 .0 4 0 .0 6 0 .0 5 - 0 .4 2

- 2 .6 5 - 0 .7 3 1 .0 6 - 1 .1 6 - 1 .8 2 - 1 .1 9 1 .4 1 1 .4 2 - 0 .5 1 0 .8 0 0 .5 9 - 5 .2 3

13 14 15 16 17 18 19 20 21 22 23 24

-0 .0 0 -0 .0 2 0 .0 0 -0 .1 1 0 .0 4 0 .0 1 -0 .0 4 -0 .0 3 -0 .0 3 0 .0 3 0 .0 7 -0 .2 7

- 0 .0 3 - 0 .2 5 0 .0 0 - 1 .4 0 0 .4 7 0 .1 2 - 0 .5 2 - 0 .3 5 - 0 .3 1 0 .3 2 0 .9 1 - 3 .4 1

25 26 27 28 29 30 31 32 33 34 35 36

0 .1 2 - 0 .0 1 - 0 .1 2 - 0 .0 7 0 .0 4 - 0 .0 4 - 0 .0 5 0 .0 2 - 0 .0 1 - 0 .0 2 0 .1 8 - 0 .2 1

1 .5 1 -0 .1 5 -1 .4 4 -0 .9 1 0 .4 4 -0 .5 3 -0 .6 0 0 .1 9 -0 .1 2 -0 .1 9 2 .2 1 -2 .5 7

37 38

0 .1 0 - 0 .0 6

1 .2 3 - 0 .7 9

wt = (1 − B12 ) (1 − B ) ln ( zt ) ‫ﻣﻦ اﻧﻤﺎط دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ‬

‫ اي ﻧﻄﺒﻖ اﻟﻨﻤﻮذج‬٠ ‫ و‬٠ ‫ اﻟﻤﻤﻜﻨﺔ هﻲ‬q ‫ و‬p ‫ﻧﺠﺪ ان ﻗﻴﻢ‬

(1 − B ) (1 − B ) ln ( z ) = (1 − ΘB ) a 12

12

t

t

‫ ﻳﻄﺒﻖ هﺬا اﻟﻨﻤﻮذج‬MINITAB ‫ اﻷﻣﺮ اﻟﺘﺎﻟﻲ ﻓﻲ‬SARIMA(٠،١،٠)(٠،١،١)١٢ ‫هﻮ‬ ARIMA 0 1 ٠ 0 1 1 12 'y(t)' ; NoConstant. zt = e y ‫ وﻟﻠﺤﺼﻮل ﻋﻠﻰ اﻟﻨﺘﺎﺋﺞ اﻟﻨﻬﺎﺋﻴﺔ ﻧﺠﺮي اﻟﺘﺤﻮﻳﻞ‬yt = ln ( zt ) ‫ﻻﺣﻆ اﻧﻨﺎ اﺳﺘﺨﺪﻣﻨﺎ‬ :‫اﻟﻨﺘﺎﺋﺞ‬ t

MTB > Name c14 = 'RESI3' c15 = 'FITS3' MTB > ARIMA 0 1 0 0 1 1 12 'y(t)' 'RESI3' 'FITS3'; SUBC> NoConstant; SUBC> Forecast 24 c7 c8 c9; SUBC> GACF; SUBC> GPACF; SUBC> GHistogram; SUBC> GNormalplot.

ARIMA Model

١٢١

ARIMA model for y(t) Estimates at each iteration Iteration SSE Parameters 0 0.0228597 0.100 1 0.0204943 0.250 2 0.0187066 0.400 3 0.0174234 0.550 4 0.0169841 0.684 5 0.0169841 0.683 6 0.0169841 0.683 Relative change in each estimate less than Final Estimates of Parameters Type Coef StDev SMA 12 0.6831 0.0610

0.0010

T 11.20

Differencing: 1 regular, 1 seasonal of order 12 Number of observations: Original series 168, after differencing 155 Residuals: SS = 0.0165799 (backforecasts excluded) MS = 0.0001077 DF = 154 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 9.0(DF=11) 29.9(DF=23) 44.5(DF=35)

48 59.4(DF=47)

Forecasts from period 168 Period 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192

Forecast 6.76750 6.70901 6.83815 6.85381 6.92288 6.89349 6.84654 6.80008 6.74395 6.75028 6.70664 6.75999 6.79052 6.73203 6.86117 6.87684 6.94590 6.91651 6.86956 6.82310 6.76697 6.77330 6.72966 6.78301

95 Percent Limits Lower Upper 6.74716 6.78784 6.68024 6.73778 6.80292 6.87338 6.81313 6.89450 6.87739 6.96836 6.84366 6.94331 6.79272 6.90035 6.74255 6.85761 6.68293 6.80497 6.68596 6.81461 6.63918 6.77410 6.68952 6.83045 6.71514 6.86590 6.65203 6.81203 6.77680 6.94554 6.78832 6.96535 6.85342 7.03838 6.82023 7.01279 6.76962 6.96950 6.71963 6.92657 6.66009 6.87385 6.66312 6.88349 6.61627 6.84305 6.66649 6.89952

(1 − B ) (1 − B ) ln ( z ) = (1 − 0.683B ) a , 12

12

t

t

Actual

‫أي أن اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح ﻟﻬﺬﻩ اﻟﻤﺘﺴﻠﺴﻠﺔ هﻮ‬

at

N ( 0, 0.0001077 )

‫ﻻﺣﻆ ان‬

Θ = 0.683, s.e. ( Θ ) = 0.061, t = 11.2

.‫أي ان اﻟﻤﻌﻠﻢ ﻋﺎﻟﻲ اﻟﻤﻌﻨﻮﻳﺔ‬ :‫ﻓﺤﺺ اﻟﺒﻮاﻗﻲ‬ ‫إﺧﺘﺒﺎر اﻟﻤﺘﻮﺳﻂ‬ MTB > ZTest 0.0 0.0103778 'RESI3'; SUBC> Alternative 0.

١٢٢

Z-Test Test of mu = 0.000000 vs mu not = 0.000000 The assumed sigma = 0.0104 Variable RESI3

N Mean StDev 155 -0.000111 0.010375

SE Mean 0.000834

Z -0.13

P 0.89

‫ اي ﻻﻧﺮﻓﺾ أن ﻣﺘﻮﺳﻂ اﻟﺒﻮاﻗﻲ ﺻﻔﺮا‬٠٫٠٥ ‫ وهﻲ اآﺒﺮ ﻣﻦ‬P-value=٠٫٨٩ ‫ﻻﺣﻆ ان اﻟـ‬ ‫إﺧﺘﺒﺎر ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬ MTB > Runs 0 'RESI3'.

Runs Test RESI3 K =

0.0000

The observed number of runs = 70 The expected number of runs = 78.1097 72 Observations above K 83 below The test is significant at 0.1893 Cannot reject at alpha = 0.05

‫ اي اﻧﻨﺎ ﻻﻧﺮﻓﺾ ﻓﺮﺿﻴﺔ ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬٠٫١٨٩٣ ‫اﻹﺧﺘﺒﺎر ﻣﻌﻨﻮي ﻋﻨﺪ‬ :‫إﺧﺘﺒﺎر إﺳﺘﻘﻼل اﻟﺒﻮاﻗﻲ‬ ‫دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬ A C F o f R e s id u a ls f o r y ( t ) ( w it h 9 5 % c o n f id e n c e l im it s f o r t h e a u t o c o r r e l a t io n s ) 1 .0 0 .8

Autocorrelation

0 .6 0 .4 0 .2 0 .0 -0 .2 -0 .4 -0 .6 -0 .8 -1 .0 3

6

9

12

15

18

21

Lag

١٢٣

24

27

30

33

36

39

P A C F o f R e s id u a ls f o r y ( t ) ( w it h 9 5 % c o n f id e n c e l im it s f o r t h e p a r t ia l a u t o c o r r e l a t io n s ) 1 .0

0 .6 0 .4 0 .2 0 .0 -0 .2 -0 .4 -0 .6 -0 .8 -1 .0 3

6

9

12

15

18

21

24

27

30

33

36

39

Lag

.‫ﻧﻼﺣﻆ اﻧﻬﺎ ﺗﻌﻄﻲ اﻧﻤﺎط اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء أي اﻧﻬﺎ ﻏﻴﺮ ﻣﺘﺮاﺑﻄﺔ وإذا آﺎﻧﺖ ﻃﺒﻴﻌﻴﺔ ﻓﻬﻲ ﻣﺴﺘﻘﻠﺔ‬ :‫إﺧﺘﺒﺎر ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬ Histogram of the Residuals (response is y(t))

Frequency

30

20

10

0 -0.03

-0.02

-0.01

0.00

0.01

0.02

0.03

0.04

Residual

Normal Probability Plot of the Residuals (response is y(t)) 0.04 0.03 0.02

Residual

Partial Autocorrelation

0 .8

0.01 0.00 -0.01 -0.02 -0.03 -3

-2

-1

0

1

2

Normal Score

١٢٤

3

‫‪K-S test for Residuals‬‬

‫‪.999‬‬ ‫‪.99‬‬ ‫‪.95‬‬

‫‪.50‬‬ ‫‪.20‬‬

‫‪Probability‬‬

‫‪.80‬‬

‫‪.05‬‬ ‫‪.01‬‬ ‫‪.001‬‬ ‫‪0.03‬‬

‫‪0.02‬‬

‫‪0.01‬‬

‫‪-0.01‬‬

‫‪0.00‬‬

‫‪-0.02‬‬

‫‪-0.03‬‬

‫‪RESI3‬‬ ‫‪Average: -0.0001115‬‬ ‫‪StDev: 0.0103754‬‬ ‫‪N: 155‬‬

‫‪Kolmogorov-Smirnov Normality Test‬‬ ‫‪D+: 0.074 D-: 0.045 D : 0.074‬‬ ‫‪Approximate P-Value: 0.041‬‬

‫ﻻﺣﻆ ان اﻟـ ‪ P-value‬ﻹﺧﺘﺒﺎر ‪ K-S‬ﻳﻌﻄﻲ ‪ ٠٫٠٤١‬وهﻲ اﻗﻞ ﻣﻦ ‪ ٠٫٠٥‬اذا اﻹﺧﺘﺒﺎر ﻣﻌﻨﻮي ﻋﻨﺪ‬ ‫‪ α = 0.05‬اي ﻻﻧﺮﻓﺾ ﻓﺮﺿﻴﺔ ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‪.‬‬ ‫اﻟﺘﻨﺒﺆ ﺑﺈﺳﺘﺨﺪام اﻟﻨﻤﻮذج‪:‬‬ ‫ﻓﻲ اﻟﻤﺨﺮﺟﺎت اﻟﺴﺎﺑﻘﺔ ﻗﻤﻨﺎ ﺑﺎﻟﺘﻨﺒﺆ ﻋﻦ ‪ ٢٤‬ﻗﻴﻤﺔ ﻣﺴﺘﻘﺒﻠﻴﺔ ﻣﻊ ‪ ٩٥%‬ﻓﺘﺮات ﺗﻨﺒﺆ وﻧﺮﺳﻤﻬﺎ ﺑﺎﻟﺮﺳﻢ‬ ‫اﻟﺘﺎﻟﻲ‪:‬‬ ‫‪1150‬‬

‫‪1050‬‬

‫‪850‬‬

‫‪750‬‬

‫‪25‬‬

‫‪20‬‬

‫‪10‬‬

‫‪15‬‬

‫‪T im e‬‬

‫واﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ ﻣﻊ اﻟﺘﻨﺒﺆات وﺣﺪود اﻟﺘﻨﺒﺆ‬

‫‪١٢٥‬‬

‫‪5‬‬

‫‪0‬‬

‫‪Forecast‬‬

‫‪950‬‬

‫‪1150‬‬

‫‪1050‬‬

‫‪850‬‬ ‫‪750‬‬

‫‪Forecast‬‬

‫‪950‬‬

‫‪650‬‬ ‫‪550‬‬ ‫‪200‬‬

‫‪0‬‬

‫‪100‬‬

‫‪Time‬‬

‫ﺣﺎﻟﺔ دراﺳﺔ ‪: ١‬‬ ‫ﺳﻮف ﻧﺤﺎول إﻳﺠﺎد ﻧﻤﻮذج ﻣﻦ ﻋﺎﺋﻠﺔ ‪ SARIMA‬ﻳﻨﻄﺒﻖ ﻋﻠﻰ اﻟﻤﺸﺎهﺪات اﻟﺘﺎﻟﻴﺔ‪:‬‬ ‫‪57.4‬‬ ‫‪60.0‬‬ ‫‪69.8‬‬ ‫‪70.1‬‬ ‫‪66.1‬‬ ‫‪58.7‬‬ ‫‪55.8‬‬ ‫‪51.6‬‬ ‫‪53.5‬‬ ‫‪53.0‬‬ ‫‪52.1‬‬ ‫‪53.6‬‬ ‫‪55.6‬‬ ‫‪62.6‬‬ ‫‪75.5‬‬ ‫‪76.4‬‬

‫‪61.5‬‬ ‫‪71.0‬‬ ‫‪71.0‬‬ ‫‪67.5‬‬ ‫‪56.1‬‬ ‫‪55.4‬‬ ‫‪55.3‬‬ ‫‪52.4‬‬ ‫‪53.4‬‬ ‫‪53.5‬‬ ‫‪53.4‬‬ ‫‪55.3‬‬ ‫‪60.1‬‬ ‫‪70.3‬‬ ‫‪73.3‬‬ ‫‪70.8‬‬

‫‪72.7‬‬ ‫‪70.6‬‬ ‫‪67.4‬‬ ‫‪58.0‬‬ ‫‪53.6‬‬ ‫‪52.9‬‬ ‫‪53.2‬‬ ‫‪52.1‬‬ ‫‪53.6‬‬ ‫‪54.3‬‬ ‫‪55.3‬‬ ‫‪58.5‬‬ ‫‪68.9‬‬ ‫‪69.4‬‬ ‫‪68.1‬‬ ‫‪60.6‬‬

‫‪72.2‬‬ ‫‪68.2‬‬ ‫‪58.2‬‬ ‫‪54.2‬‬ ‫‪53.0‬‬ ‫‪53.0‬‬ ‫‪52.8‬‬ ‫‪52.6‬‬ ‫‪55.1‬‬ ‫‪56.3‬‬ ‫‪58.2‬‬ ‫‪69.3‬‬ ‫‪68.8‬‬ ‫‪64.7‬‬ ‫‪58.1‬‬ ‫‪56.4‬‬

‫‪71.5‬‬ ‫‪57.7‬‬ ‫‪54.3‬‬ ‫‪53.2‬‬ ‫‪52.8‬‬ ‫‪52.7‬‬ ‫‪52.8‬‬ ‫‪53.9‬‬ ‫‪57.3‬‬ ‫‪59.4‬‬ ‫‪66.9‬‬ ‫‪69.6‬‬ ‫‪63.6‬‬ ‫‪57.1‬‬ ‫‪54.8‬‬ ‫‪55.6‬‬

‫‪59.1‬‬ ‫‪54.6‬‬ ‫‪53.4‬‬ ‫‪53.0‬‬ ‫‪52.6‬‬ ‫‪53.4‬‬ ‫‪54.5‬‬ ‫‪56.4‬‬ ‫‪61.9‬‬ ‫‪70.4‬‬ ‫‪70.7‬‬ ‫‪64.2‬‬ ‫‪57.1‬‬ ‫‪53.9‬‬ ‫‪54.3‬‬ ‫‪55.0‬‬

‫‪57.2‬‬ ‫‪54.9‬‬ ‫‪53.0‬‬ ‫‪53.0‬‬ ‫‪52.9‬‬ ‫‪54.9‬‬ ‫‪56.4‬‬ ‫‪61.7‬‬ ‫‪69.9‬‬ ‫‪69.9‬‬ ‫‪65.3‬‬ ‫‪55.5‬‬ ‫‪52.2‬‬ ‫‪53.5‬‬ ‫‪54.6‬‬ ‫‪54.7‬‬

‫‪56.3‬‬ ‫‪54.9‬‬ ‫‪52.8‬‬ ‫‪53.4‬‬ ‫‪54.0‬‬ ‫‪57.5‬‬ ‫‪59.3‬‬ ‫‪68.3‬‬ ‫‪71.5‬‬ ‫‪61.0‬‬ ‫‪56.5‬‬ ‫‪53.3‬‬ ‫‪51.5‬‬ ‫‪53.5‬‬ ‫‪54.2‬‬ ‫‪55.8‬‬

‫‪55.7‬‬ ‫‪55.3‬‬ ‫‪54.4‬‬ ‫‪55.6‬‬ ‫‪56.6‬‬ ‫‪69.3‬‬ ‫‪70.0‬‬ ‫‪65.3‬‬ ‫‪57.3‬‬ ‫‪53.9‬‬ ‫‪52.5‬‬ ‫‪51.5‬‬ ‫‪51.5‬‬ ‫‪53.3‬‬ ‫‪55.8‬‬ ‫‪60.5‬‬ ‫‪62.2‬‬

‫‪55.8‬‬ ‫‪54.9‬‬ ‫‪53.3‬‬ ‫‪54.6‬‬ ‫‪54.9‬‬ ‫‪61.2‬‬ ‫‪68.7‬‬ ‫‪67.9‬‬ ‫‪65.1‬‬ ‫‪52.7‬‬ ‫‪53.4‬‬ ‫‪52.4‬‬ ‫‪51.7‬‬ ‫‪53.1‬‬ ‫‪54.8‬‬ ‫‪57.7‬‬

‫اﻟﻤﺨﻄﻂ اﻟﺰﻣﻨﻲ‬

‫‪7 0‬‬

‫)‪z(t‬‬ ‫‪6 0‬‬

‫‪5 0‬‬ ‫‪1 5 0‬‬

‫‪5 0‬‬

‫‪1 0 0‬‬

‫‪١٢٦‬‬

‫‪In d e x‬‬

‫)‪z(t‬‬ ‫‪56.3‬‬ ‫‪56.9‬‬ ‫‪56.0‬‬ ‫‪59.4‬‬ ‫‪68.2‬‬ ‫‪69.8‬‬ ‫‪67.9‬‬ ‫‪58.2‬‬ ‫‪52.7‬‬ ‫‪53.3‬‬ ‫‪53.2‬‬ ‫‪51.5‬‬ ‫‪52.3‬‬ ‫‪53.9‬‬ ‫‪57.9‬‬ ‫‪66.4‬‬ ‫‪74.8‬‬

‫ ﻧﺠﺪ‬wt = (1 − B ) zt ‫ﺑﺄﺧﺬ اﻟﻔﺮق اﻷول ﻹﺳﺘﻘﺮار اﻟﻤﺘﻮﺳﻂ‬

y(t)

1 0

0

-1 0

5 0

In d e x

1 0 0

1 5 0

‫وﻟﻬﺎ دوال ﺗﺮاﺑﻂ ذاﺗﻲ وﺗﺮاﺑﻂ ذاﺗﻲ ﺟﺰﺋﻲ‬ Autocorrelation

Autocorrelation Function for y(t) 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

Lag

Corr

T

LBQ

Lag

Corr

T

LBQ

Lag

Corr

T

LBQ

Lag

Corr

T

LBQ

1 2 3 4 5 6 7 8 9 10 11 12

0.47 -0.02 -0.37 -0.28 -0.21 -0.19 -0.21 -0.27 -0.32 -0.00 0.46 0.86

6.32 -0.22 -4.10 -2.82 -2.00 -1.77 -1.92 -2.46 -2.85 -0.02 3.91 6.71

40.60 40.67 65.83 80.10 87.92 94.41 102.29 115.89 135.47 135.48 176.37 318.37

13 14 15 16 17 18 19 20 21 22 23 24

0.43 -0.01 -0.33 -0.25 -0.19 -0.17 -0.19 -0.25 -0.31 -0.01 0.42 0.79

2.76 -0.03 -2.00 -1.49 -1.12 -0.97 -1.09 -1.42 -1.76 -0.06 2.35 4.28

354.69 354.69 375.77 388.04 395.15 400.68 407.71 419.89 439.21 439.24 475.10 602.75

25 26 27 28 29 30 31 32 33 34 35 36

0.41 -0.00 -0.29 -0.23 -0.17 -0.16 -0.17 -0.23 -0.28 -0.01 0.37 0.72

2.06 -0.01 -1.43 -1.08 -0.82 -0.78 -0.83 -1.06 -1.31 -0.06 1.70 3.27

638.60 638.60 656.89 667.77 674.14 679.93 686.56 697.63 714.89 714.93 745.11 860.42

37 38 39 40 41 42 43 44

0.40 0.02 -0.27 -0.21 -0.17 -0.15 -0.16 -0.20

1.72 0.07 -1.14 -0.89 -0.71 -0.61 -0.65 -0.83

896.48 896.54 913.29 923.75 930.47 935.55 941.37 950.82

2

12

22

32

42

Partial Autocorrelation

Partial Autocorrelation Function for y(t) 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

2

12

Lag PAC

T

0.47 -0.32 -0.30 0.09 -0.23 -0.25 -0.16 -0.43 -0.64 -0.31 -0.16 0.41

6.32 -4.21 -4.00 1.22 -3.05 -3.33 -2.09 -5.67 -8.56 -4.11 -2.18 5.42

1 2 3 4 5 6 7 8 9 10 11 12

Lag PAC 13 14 15 16 17 18 19 20 21 22 23 24

-0.27 -0.01 0.01 -0.07 -0.01 0.02 0.03 0.07 -0.12 -0.02 -0.06 0.13

22 T -3.62 -0.17 0.13 -0.95 -0.16 0.24 0.44 0.89 -1.62 -0.26 -0.86 1.74

32

Lag PAC 25 26 27 28 29 30 31 32 33 34 35 36

-0.02 -0.06 0.07 -0.00 -0.02 -0.04 0.01 0.03 -0.02 0.00 -0.09 0.05

T -0.33 -0.74 0.99 -0.05 -0.26 -0.60 0.20 0.37 -0.25 0.06 -1.17 0.61

42

Lag PAC 37 38 39 40 41 42 43 44

0.03 -0.03 -0.01 0.02 -0.04 0.03 -0.00 0.03

T 0.38 -0.35 -0.17 0.23 -0.56 0.35 -0.00 0.43

‫ وﻳﻨﺘﺞ اﻟﻤﺘﺴﻠﺴﻠﺔ‬wt = (1 − B12 ) (1 − B ) zt ‫ أي‬١٢ ‫ﻧﺮى اﻧﻬﺎ ﺗﺤﺘﺎج إﻟﻰ ﺗﻔﺮﻳﻖ ﻣﻮﺳﻤﻲ ﻣﻦ اﻟﺮﺗﺒﺔ‬ ‫اﻟﺘﺎﻟﻴﺔ‬

١٢٧

w(t)

5

0

-5 In d e x

50

100

150

‫وﻟﻬﺎ دوال ﺗﺮاﺑﻂ ذاﺗﻲ وﺗﺮاﺑﻂ ذاﺗﻲ ﺟﺰﺋﻲ‬ Autocorrelation

A u t o c o r r e la t io n F u n c t io n f o r w ( t ) 1 .0 0 .8 0 .6 0 .4 0 .2 0 .0 - 0 .2 - 0 .4 - 0 .6 - 0 .8 - 1 .0

10

20

30

L ag

C o rr

T

LBQ

L ag

C o rr

T

LBQ

L ag

C o rr

T

LBQ

1 2 3 4 5 6 7 8 9 10 11 12

-0 .0 1 -0 .1 9 -0 .2 4 -0 .0 0 0 .0 0 -0 .0 4 -0 .0 4 -0 .0 3 0 .2 2 0 .1 0 0 .0 8 -0 .4 2

- 0 .1 9 - 2 .4 8 - 2 .9 7 - 0 .0 3 0 .0 3 - 0 .5 0 - 0 .4 9 - 0 .3 1 2 .5 6 1 .1 6 0 .9 3 - 4 .6 3

0 .0 4 6 .3 1 1 6 .1 0 1 6 .1 0 1 6 .1 0 1 6 .4 1 1 6 .7 1 1 6 .8 3 2 5 .2 6 2 7 .1 3 2 8 .3 7 5 9 .4 0

13 14 15 16 17 18 19 20 21 22 23 24

-0 .0 8 0 .1 0 0 .1 2 -0 .0 2 -0 .0 3 0 .1 0 0 .0 7 -0 .0 1 -0 .2 0 -0 .0 2 0 .1 0 0 .1 4

-0 .8 4 0 .9 8 1 .1 5 -0 .1 7 -0 .2 8 1 .0 1 0 .7 0 -0 .0 5 -1 .9 4 -0 .2 3 0 .9 5 1 .3 0

6 0 .6 9 6 2 .5 0 6 5 .0 1 6 5 .0 7 6 5 .2 2 6 7 .2 4 6 8 .2 3 6 8 .2 4 7 5 .9 4 7 6 .0 5 7 8 .0 2 8 1 .7 8

25 26 27 28 29 30 31 32 33 34 35 36

- 0 .0 8 - 0 .1 3 - 0 .0 1 0 .0 6 0 .1 2 - 0 .1 5 - 0 .0 8 - 0 .0 3 0 .1 9 0 .1 7 - 0 .1 0 - 0 .2 2

-0 .7 6 -1 .1 8 -0 .1 3 0 .5 6 1 .0 9 -1 .3 6 -0 .7 2 -0 .2 8 1 .6 6 1 .4 9 -0 .8 6 -1 .9 2

8 3 .0 9 8 6 .2 8 8 6 .3 3 8 7 .0 7 8 9 .9 4 9 4 .4 7 9 5 .8 0 9 6 .0 1 1 0 3 .2 0 1 0 9 .1 9 1 1 1 .2 4 1 2 1 .7 7

40

L ag

T

LBQ

3 7 0 .0 5 0 .4 6 3 8 0 .1 0 0 .8 4 3 9 0 .0 2 0 .1 5 4 0 -0 .0 2 - 0 .1 6 4 1 -0 .1 2 - 1 .0 1

C o rr

1 2 2 .4 0 1 2 4 .5 2 1 2 4 .5 9 1 2 4 .6 7 1 2 7 .8 4

Partial Autocorrelation

P a r t ia l A u t o c o r r e la t io n F u n c t io n f o r w ( t ) 1 .0 0 .8 0 .6 0 .4 0 .2 0 .0 -0 .2 -0 .4 -0 .6 -0 .8 -1 .0

10

20

30

40

L ag

PAC

T

L ag

PAC

T

L ag

PAC

T

L ag

PAC

T

1 2 3 4 5 6 7 8 9 10 11 12

- 0 .0 1 - 0 .1 9 - 0 .2 6 - 0 .0 7 - 0 .1 1 - 0 .1 4 - 0 .1 1 - 0 .1 3 0 .1 4 0 .0 7 0 .1 7 - 0 .3 2

- 0 .1 9 - 2 .4 8 - 3 .2 8 - 0 .8 5 - 1 .4 3 - 1 .8 4 - 1 .4 6 - 1 .6 6 1 .7 6 0 .9 5 2 .1 9 - 4 .0 8

13 14 15 16 17 18 19 20 21 22 23 24

-0 .0 5 0 .0 1 -0 .0 4 -0 .0 3 -0 .0 2 0 .0 6 0 .0 5 -0 .0 3 -0 .0 5 0 .0 5 0 .1 9 0 .0 1

- 0 .6 9 0 .1 5 - 0 .5 0 - 0 .3 3 - 0 .2 3 0 .7 8 0 .6 2 - 0 .4 2 - 0 .6 8 0 .6 3 2 .5 0 0 .0 9

25 26 27 28 29 30 31 32 33 34 35 36

- 0 .1 2 - 0 .0 8 - 0 .0 3 - 0 .0 3 0 .0 7 - 0 .1 1 - 0 .0 2 - 0 .1 0 - 0 .0 1 0 .1 8 0 .0 7 - 0 .0 9

-1 .4 9 -0 .9 7 -0 .3 2 -0 .3 6 0 .8 7 -1 .3 9 -0 .2 9 -1 .2 9 -0 .0 9 2 .3 1 0 .9 5 -1 .1 8

37 38 39 40 41

0 .0 2 - 0 .0 8 0 .0 3 0 .0 5 - 0 .0 3

0 .2 4 - 1 .0 6 0 .4 4 0 .6 3 - 0 .3 6

‫ﻣﻦ اﻷﻧﻤﺎط اﻟﻤﺸﺎهﺪة ﻟﺪوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻗﺪ ﻳﻜﻮن اﻟﻨﻤﻮذج اﻟﻤﻨﺎﺳﺐ هﻮ‬ ‫ أي‬SARIMA(١،١،١)(٠،١،١)١٢

١٢٨

(1 − φ B ) (1 − B12 ) (1 − B ) zt = (1 − θ B ) (1 − ΘB12 ) at :‫ﻧﻄﺒﻖ هﺬا اﻟﻨﻤﻮذج ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ آﺎﻟﺘﺎﻟﻲ‬ MTB > ARIMA 1 1 1 0 1 1 12 'z(t)' 'RESI2'; SUBC> NoConstant; SUBC> GACF; SUBC> GPACF; SUBC> GHistogram; SUBC> GNormalplot.

ARIMA Model ARIMA model for z(t) Estimates at each iteration Iteration SSE Parameters 0 307.653 0.100 0.100 1 281.217 0.100 0.100 2 262.275 0.226 0.231 3 262.027 0.376 0.381 4 261.770 0.526 0.531 5 261.426 0.675 0.681 6 260.905 0.824 0.831 7 260.036 0.970 0.981 8 227.926 0.835 0.980 9 221.838 0.748 0.980 10 221.665 0.738 0.980 11 221.637 0.738 0.980 12 221.610 0.737 0.980 13 221.585 0.737 0.980 Relative change in each estimate less than Final Estimates of Parameters Type Coef StDev AR 1 0.7374 0.0620 MA 1 0.9796 0.0017 SMA 12 0.5898 0.0736

0.100 0.250 0.400 0.401 0.401 0.402 0.403 0.405 0.536 0.576 0.586 0.589 0.589 0.590 0.0010

T 11.89 582.86 8.01

Differencing: 1 regular, 1 seasonal of order 12 Number of observations: Original series 178, after differencing 165 Residuals: SS = 214.393 (backforecasts excluded) MS = 1.323 DF = 162 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 15.7(DF= 9) 30.9(DF=21) 61.6(DF=33)

48 67.1(DF=45)

‫أي أن اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح ﻟﻬﺬﻩ اﻟﻤﺘﺴﻠﺴﻠﺔ هﻮ‬

(1 − 0.74 B ) (1 − B ) (1 − B ) zt = (1 − 0.98B ) (1 − 0.59 B12 ) at , 12

at

N ( 0,1.323)

‫ﻻﺣﻆ ان‬

φ = 0.74, s.e. (φ ) = 0.062, t = 11.89 θ = 0.96, s.e. (θ ) = 0.0017, t = 582.86 Θ = 0.59, s.e. ( Θ ) = 0.074, t = 8.01

.‫أي ان اﻟﻤﻌﺎﻟﻢ ﻋﺎﻟﻴﺔ اﻟﻤﻌﻨﻮﻳﺔ‬ :‫ﻓﺤﺺ اﻟﺒﻮاﻗﻲ‬ ‫إﺧﺘﺒﺎر اﻟﻤﺘﻮﺳﻂ‬ MTB > ZTest 0.0 1.15 'RESI1'; SUBC> Alternative 0.

Z-Test

١٢٩

Test of mu = 0.0000 vs mu not = 0.0000 The assumed sigma = 1.15 Variable RESI1

N 165

Mean -0.0144

StDev 1.1433

SE Mean 0.0895

Z -0.16

P 0.87

‫ اي ﻻﻧﺮﻓﺾ أن ﻣﺘﻮﺳﻂ اﻟﺒﻮاﻗﻲ ﺻﻔﺮا‬٠٫٠٥ ‫ وهﻲ اآﺒﺮ ﻣﻦ‬P-value=٠٫٨٧ ‫ﻻﺣﻆ ان اﻟـ‬ ‫إﺧﺘﺒﺎر ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬ MTB > Runs 0 'RESI1'.

Runs Test RESI1 K =

0.0000

The observed number of runs = 67 The expected number of runs = 82.4061 73 Observations above K 92 below The test is significant at 0.0149

‫ اي اﻧﻨﺎ ﻧﺮﻓﺾ ﻓﺮﺿﻴﺔ ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ وهﺬا ﻳﺤﺘﺎج إﻟﻰ إﺟﺮاء‬٠٫٠٥ ‫اﻹﺧﺘﺒﺎر ﻏﻴﺮﻣﻌﻨﻮي ﻋﻨﺪ‬ :‫ ﻋﻠﻰ اﻟﻮﺳﻴﻂ اﻟﺘﺎﻟﻲ‬Sign Test ‫إﺧﺘﺒﺎر ﺁﺧﺮ أآﺜﺮ ﻗﻮة ﻣﻦ إﺧﺘﺒﺎر اﻟﺠﺮي ﻣﺜﻞ إﺧﺘﺒﺎر اﻹﺷﺎرة‬ MTB > STest 0.0 'RESI1'; SUBC> Alternative 0.

Sign Test for Median Sign test of median = 0.00000 versus RESI1

N 165

N* 13

Below 92

Equal 0

not = Above 73

0.00000 P 0.1611

Median -0.08139

‫ وﻟﻠﺘﺄآﺪ ﻧﺠﺮي إﺧﺘﺒﺎر وﻟﻜﻮآﺴﻮن ﻹﺷﺎرات اﻟﺮﺗﺐ ﻋﻠﻰ اﻟﻮﺳﻴﻂ‬٠٫١٦١١ ‫واﻹﺧﺘﺒﺎر ﻣﻌﻨﻮي ﻋﻨﺪ‬ ‫اﻟﺘﺎﻟﻲ‬ MTB > WTest 0.0 'RESI1'; SUBC> Alternative 0.

Wilcoxon Signed Rank Test Test of median = 0.000000 versus median not = 0.000000

RESI1

N 165

Number Missing 13

N for Test 165

Wilcoxon Statistic 6321.0

P 0.392

Estimated Median -0.05940

٠٫٣٩٢ ‫واﻹﺧﺘﺒﺎر اﻳﻀﺎ ﻣﻌﻨﻮي ﻋﻨﺪ‬ :‫إﺧﺘﺒﺎر إﺳﺘﻘﻼل اﻟﺒﻮاﻗﻲ‬ ‫دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬

١٣٠

A C F o f R e s id u a ls f o r z ( t) ( w it h 9 5 % c o n f id e n c e l im it s f o r t h e a u t o c o r r e l a t io n s ) 1 .0 0 .8

Autocorrelation

0 .6 0 .4 0 .2 0 .0 -0 .2 -0 .4 -0 .6 -0 .8 -1 .0

3

6

9

12

15

18

21

24

27

30

33

36

39

36

39

Lag

P A C F o f R e s id u a ls f o r z ( t) ( w it h 9 5 % c o n f id e n c e l im it s f o r t h e p a r t ia l a u t o c o r r e l a t io n s ) 1 .0

Partial Autocorrelation

0 .8 0 .6 0 .4 0 .2 0 .0 -0 .2 -0 .4 -0 .6 -0 .8 -1 .0

3

6

9

12

15

18

21

24

27

30

33

Lag

.‫ﻧﻼﺣﻆ اﻧﻬﺎ ﺗﻌﻄﻲ اﻧﻤﺎط اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء أي اﻧﻬﺎ ﻏﻴﺮ ﻣﺘﺮاﺑﻄﺔ وإذا آﺎﻧﺖ ﻃﺒﻴﻌﻴﺔ ﻓﻬﻲ ﻣﺴﺘﻘﻠﺔ‬ :‫إﺧﺘﺒﺎر ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬ Histogram of the Residuals (response is z(t)) 50

Frequency

40

30

20

10

0 -6

-5

-4

-3

-2

-1

0

1

2

Residual

١٣١

3

4

Normal Probability Plot of the Residuals (response is z(t)) 4 3 2

Residual

1 0 -1 -2 -3 -4 -5 -3

-2

-1

0

1

2

3

Normal Score

K-S Test for Residuals

.999

Probability

.99 .95 .80 .50 .20 .05 .01 .001 -5

-4

-3

-2

-1

0

1

2

3

RESI1 Average: -0.0144171 StDev: 1.14327 N: 165

Kolmogorov-Smirnov Normality Test D+: 0.117 D-: 0.140 D : 0.140 Approximate P-Value < 0.01

‫ اي‬α = 0.05 ‫ اذا اﻹﺧﺘﺒﺎر ﻣﻌﻨﻮي ﻋﻨﺪ‬٠٫٠١ ‫ أﻗﻞ ﻣﻦ‬K-S ‫ ﻹﺧﺘﺒﺎر‬P-value ‫ﻻﺣﻆ ان اﻟـ‬ .‫ﻻﻧﺮﻓﺾ ﻓﺮﺿﻴﺔ ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬ :‫اﻟﺘﻨﺒﺆ ﺑﺈﺳﺘﺨﺪام اﻟﻨﻤﻮذج‬ ‫ ﻓﺘﺮات ﺗﻨﺒﺆ‬٩٥% ‫ ﻗﻴﻤﺔ ﻣﺴﺘﻘﺒﻠﻴﺔ ﻣﻊ‬٣٦ ‫ﺳﻨﻘﻮم ﺑﺎﻟﺘﻨﺒﺆ ﻋﻦ‬ Forecasts from period 178 Period 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193

Forecast 57.7885 55.8516 54.7429 54.1820 54.6298 54.9152 55.6640 59.3778 68.6924 74.0698 73.9650 63.5866 58.9095 56.7768 55.5237

95 Percent Limits Lower Upper 55.5332 60.0437 53.0220 58.6812 51.6264 57.8594 50.9063 57.4578 51.2601 57.9994 51.4875 58.3430 52.1986 59.1294 55.8869 62.8688 65.1833 72.2015 70.5472 77.5925 70.4317 77.4983 60.0447 67.1286 55.1843 62.6347 52.9407 60.6128 51.6169 59.4305

١٣٢

Actual

‫‪50.9022‬‬ ‫‪51.2380‬‬ ‫‪51.4409‬‬ ‫‪52.1278‬‬ ‫‪55.7947‬‬ ‫‪65.0729‬‬ ‫‪70.4217‬‬ ‫‪70.2940‬‬ ‫‪59.8968‬‬ ‫‪55.0418‬‬ ‫‪52.7962‬‬ ‫‪51.4676‬‬ ‫‪50.7472‬‬ ‫‪51.0774‬‬ ‫‪51.2749‬‬ ‫‪51.9568‬‬ ‫‪55.6189‬‬ ‫‪64.8927‬‬ ‫‪70.2374‬‬ ‫‪70.1057‬‬ ‫‪59.7046‬‬

‫‪58.8105‬‬ ‫‪59.2131‬‬ ‫‪59.4653‬‬ ‫‪60.1905‬‬ ‫‪63.8885‬‬ ‫‪73.1929‬‬ ‫‪78.5647‬‬ ‫‪78.4575‬‬ ‫‪68.0793‬‬ ‫‪63.5664‬‬ ‫‪61.5364‬‬ ‫‪60.3514‬‬ ‫‪59.7315‬‬ ‫‪60.1357‬‬ ‫‪60.3903‬‬ ‫‪61.1184‬‬ ‫‪64.8194‬‬ ‫‪74.1270‬‬ ‫‪79.5021‬‬ ‫‪79.3983‬‬ ‫‪69.0235‬‬

‫‪194‬‬ ‫‪195‬‬ ‫‪196‬‬ ‫‪197‬‬ ‫‪198‬‬ ‫‪199‬‬ ‫‪200‬‬ ‫‪201‬‬ ‫‪202‬‬ ‫‪203‬‬ ‫‪204‬‬ ‫‪205‬‬ ‫‪206‬‬ ‫‪207‬‬ ‫‪208‬‬ ‫‪209‬‬ ‫‪210‬‬ ‫‪211‬‬ ‫‪212‬‬ ‫‪213‬‬ ‫‪214‬‬

‫‪54.8564‬‬ ‫‪55.2256‬‬ ‫‪55.4531‬‬ ‫‪56.1592‬‬ ‫‪59.8416‬‬ ‫‪69.1329‬‬ ‫‪74.4932‬‬ ‫‪74.3758‬‬ ‫‪63.9881‬‬ ‫‪59.3041‬‬ ‫‪57.1663‬‬ ‫‪55.9095‬‬ ‫‪55.2394‬‬ ‫‪55.6066‬‬ ‫‪55.8326‬‬ ‫‪56.5376‬‬ ‫‪60.2191‬‬ ‫‪69.5099‬‬ ‫‪74.8697‬‬ ‫‪74.7520‬‬ ‫‪64.3640‬‬

‫وﻧﺮﺳﻤﻬﺎ ﺑﺎﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ‪:‬‬ ‫‪8 0‬‬

‫‪Forecast‬‬

‫‪7 0‬‬

‫‪6 0‬‬

‫‪5 0‬‬

‫‪3 0‬‬

‫‪4 0‬‬

‫‪2 0‬‬

‫‪0‬‬

‫‪1 0‬‬

‫‪T im e‬‬

‫اﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ ﺑﻜﺎﻣﻠﻬﺎ ﻣﻊ اﻟﺘﻨﺒﺆات وﻓﺘﺮات اﻟﺘﻨﺒﺆ‬

‫‪80‬‬

‫‪60‬‬

‫‪50‬‬

‫‪200‬‬

‫‪100‬‬

‫‪T im e‬‬

‫‪١٣٣‬‬

‫‪0‬‬

‫‪Forecast‬‬

‫‪70‬‬

١٣٤

‫اﻟﻔﺼﻞ اﻟﺴﺎﺑﻊ‬ ‫ورﻗﺔ ﺗﺪرﻳﺐ ﻋﻤﻠﻲ ﻋﻠﻲ اﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ ﻧﻤﺎذج اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك‪-‬اﻹﻧﺤﺪار‬ ‫اﻟﺬاﺗﻲ‬ ‫‪Forecasting By ARMA Models‬‬ ‫اﻟﻤﺸﺎهﺪات اﻟﺘﺎﻟﻴﺔ ﻟﻈﺎهﺮة ﻋﺸﻮاﺋﻴﺔ ﻣﺴﺠﻠﺔ ﻋﻠﻲ ﺷﻜﻞ ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ‬ ‫‪28.0‬‬ ‫‪28.0‬‬ ‫‪29.5‬‬

‫‪23.5‬‬ ‫‪30.5‬‬ ‫‪32.3‬‬

‫‪21.3‬‬ ‫‪29.6‬‬ ‫‪30.6‬‬

‫‪24.5‬‬ ‫‪36.5‬‬ ‫‪26.5‬‬

‫‪23.5‬‬ ‫‪36.5‬‬ ‫‪20.7‬‬ ‫‪16.4‬‬

‫‪15.3‬‬ ‫‪25.0‬‬ ‫‪16.0‬‬ ‫‪23.4‬‬

‫‪15.5‬‬ ‫‪25.3‬‬ ‫‪19.0‬‬ ‫‪26.4‬‬

‫‪21.0‬‬ ‫‪17.3‬‬ ‫‪19.7‬‬ ‫‪32.2‬‬

‫‪12.0‬‬ ‫‪24.0‬‬ ‫‪26.0‬‬ ‫‪28.3‬‬

‫‪20.5‬‬ ‫‪15.5‬‬ ‫‪21.5‬‬ ‫‪31.3‬‬

‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‪:‬‬ ‫;‪MTB > TSPlot C1‬‬ ‫>‪SUBC‬‬ ‫;‪Index‬‬ ‫>‪SUBC‬‬ ‫;‪TDisplay 11‬‬ ‫>‪SUBC‬‬ ‫;‪Symbol‬‬ ‫>‪SUBC‬‬ ‫‪Connect.‬‬

‫‪30‬‬

‫‪C‬‬ ‫‪1‬‬ ‫‪20‬‬

‫‪10‬‬ ‫‪30‬‬

‫‪10‬‬

‫‪20‬‬

‫اوﻻ ﻧﻮﺟﺪ دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬

‫‪١٣٥‬‬

‫‪Index‬‬

MTB > %ACF C1.

Autocorrelation

Autocorrelation Function for C1 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

1

2

3

4

Lag

Corr

T

LBQ

1 2 3 4 5 6 7

0.63 0.30 0.14 -0.05 -0.24 -0.30 -0.24

3.79 1.35 0.59 -0.20 -1.01 -1.22 -0.94

15.62 19.27 20.07 20.17 22.71 26.71 29.40

5

Lag

6

Corr

7

T

LBQ

8 -0.20 -0.78 9 -0.12 -0.44

31.42 32.10

8

9

MTB > %PACF C1. Executing from file: E:\MTBWIN\MACROS\PACF.MAC

Partial Autocorrelation

Partial Autocorrelation Function for C1 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

1

2

3

4

Lag PAC 1 2 3 4 5 6 7

0.63 -0.16 0.04 -0.20 -0.18 -0.04 0.02

5

6

T

Lag PAC

T

3.79 -0.98 0.22 -1.20 -1.09 -0.23 0.12

8 -0.07 9 0.05

-0.39 0.29

7

8

9

ARMA (1,1) ‫ﻧﻼﺣﻆ ﻣﻦ اﻧﻤﺎط اﻟﺪاﻟﺘﻴﻦ ان اﻟﻤﺸﺎهﺪات ﻗﺪ ﺗﻜﻮن ﻣﻦ ﻧﻤﻮذج‬ ‫ﻧﻄﺒﻖ اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح‬ MTB > Name c17 = 'RESI1' MTB > ARIMA 1 0 1 C1 'RESI1'; SUBC> Constant; SUBC> Forecast 5 c14 c15 c16; SUBC> GACF; SUBC> GPACF; SUBC> GNormalplot.

١٣٦

ARIMA Model ARIMA model for C1 Estimates at each iteration Iteration SSE Parameters 0 1337.71 0.100 0.100 21.918 1 936.95 0.250 -0.049 18.193 2 849.78 0.211 -0.199 19.106 3 751.53 0.215 -0.349 18.941 4 658.66 0.266 -0.499 17.594 5 592.30 0.372 -0.649 14.890 6 580.80 0.433 -0.699 13.314 7 579.30 0.455 -0.714 12.698 8 579.11 0.464 -0.719 12.470 9 579.08 0.467 -0.721 12.386 10 579.08 0.468 -0.722 12.356 11 579.08 0.468 -0.722 12.345 Relative change in each estimate less than 0.0010 Final Estimates of Parameters Type Coef StDev AR 1 0.4684 0.1755 MA 1 -0.7221 0.1380 Constant 12.345 1.154 Mean 23.221 2.170

T 2.67 -5.23 10.70

Number of observations: 36 Residuals: SS = 523.365 (backforecasts excluded) MS = 15.860 DF = 33 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 48 Chi-Square 7.2(DF=10) 15.9(DF=22) * (DF= *) * (DF= *) Forecasts from period 36 Period 37 38 39 40 41

Forecast 14.7649 19.2606 21.3663 22.3524 22.8143

95 Percent Limits Lower Upper 6.9578 22.5720 7.1228 31.3985 8.4715 34.2610 9.2975 35.4074 9.7245 35.9041

zt = 12.345 + 0.4684 zt −1 + at − 0.7221at −1 , at ‫ﺣﻴﺚ‬ φˆ1 = 0.4684 se φˆ1 = 0.1755 t = 2.67

Actual

:‫اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح هﻮ‬ WN ( 0,15.86 ) ∀t

( ) θˆ = −0.7221 se (θˆ ) = 0.1380 t = −5.23 δˆ = 12.345 se (δˆ ) = 1.154 t = 10.70 1

1

σˆ 2 = 15.86 df = 33 ‫ ﻓﻤﺜﻼ اﻟﻔﺮﺿﻴﺔ‬α = 0.05 ‫وﻧﻼﺣﻆ ان ﺟﻤﻴﻊ اﻟﻤﻘﺪرات ﻣﻌﻨﻮﻳﺔ ﻋﻨﺪ‬

١٣٧

H 0 : φ1 = 0 H1 : φ1 ≠ 0 ‫ أي‬α = 0.05 ‫ وهﻲ ﻣﻌﻨﻮﻳﺔ ﻋﻨﺪ‬t =

0.4684 φˆ1 = = 2.6689 ‫ﻧﺨﺘﺒﺮهﺎ ﺑﺎﻹﺣﺼﺎﺋﺔ‬ 0.1755 se φˆ1

( )

‫ وﺑﺎﻟﻤﺜﻞ ﻟﺠﻤﻴﻊ اﻟﻘﺪرات اﻻﺧﺮى‬φ1 = 0 ‫اﻧﻨﺎ ﻧﺮﻓﺾ ان‬

‫ﺛﺎﻧﻴﺎ ﻓﺤﺺ اﻟﺒﻮاﻗﻲ‬ :‫إﺧﺘﺒﺎر ﻣﺘﻮﺳﻂ اﻟﺒﻮاﻗﻲ‬ ‫اﻹﺧﺘﺒﺎر هﻮ‬

H 0 : µa = 0, H1 : µa ≠ 0

MTB > TTest 0.0 'RESI1'; SUBC> Alternative 0.

T-Test of the Mean Test of mu = 0.000 vs mu not = 0.000 Variable RESI1

N 36

Mean 0.344

StDev 3.851

SE Mean 0.642

T 0.54

P 0.60

‫ أي ان اﻹﺧﺘﺒﺎر‬α = 0.05 ‫ وهﻲ اآﺒﺮ ﻣﻦ‬0.6 ‫ ﻟﻬﺎ هﻲ‬P-Value ‫ واﻟـ‬t = 0.54 ‫ﻻﺣﻆ ان‬ ‫ﻏﻴﺮ ﻣﻌﻨﻮي أي ﻳﻤﻜﻦ إﻋﺘﺒﺎر ﻣﺘﻮﺳﻂ اﻟﺒﻮاﻗﻲ ﻳﺴﺎوي اﻟﺼﻔﺮ‬ :‫إﺧﺘﺒﺎرﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬ ‫وﻧﺴﺘﺨﺪم ﻟﺬﻟﻚ إﺧﺘﺒﺎر اﻟﺠﺮي‬Runs Test MTB > Runs 'RESI1'.

Runs Test RESI1 K =

0.3443

The observed number of runs = 21 The expected number of runs = 19.0000 18 Observations above K 18 below The test is significant at 0.4989 Cannot reject at alpha = 0.05

α = 0.05 ‫ﻻﻳﻤﻜﻨﻨﺎ رﻓﺾ ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ ﻋﻨﺪ‬ :‫إﺧﺘﺒﺎر ﺗﺮاﺑﻂ اﻟﺒﻮاﻗﻲ‬ ‫وﻧﺴﺘﺨﺪم ﻟﺬﻟﻚ إﺧﺘﺒﺎر اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ‬

ACF of Residuals for C1 (with 95% confidence limits for the autocorrelations) 1.0 0.8

orrelation

0.6 0.4 0.2 0.0

١٣٨

PACF of Residuals for C1 (with 95% confidence limits for the partial autocorrelations) 1.0

Partial Autocorrelation

0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0 1

2

3

4

5

6

7

8

9

Lag

‫ﻧﻼﺣﻆ اﻧﻪ ﻻﻳﻮﺟﺪ أي ﺗﺮاﺑﻂ ﻣﻦ أي درﺟﺔ ﺑﻴﻦ اﻟﻘﻴﻢ اﻟﻤﺨﺘﻠﻔﺔ ﻟﻠﺒﻮاﻗﻲ أي اﻧﻬﺎ ﺗﻈﻬﺮ اﻧﻤﺎط ﺗﺘﻤﺸﻲ‬ ‫ﻣﻊ آﻮﻧﻬﺎ ﻣﺘﺴﻠﺴﺔ ﺿﺠﺔ ﺑﻴﻀﺎء‬ Normal Probability Plot ‫واﺧﻴﺮا ﻧﺨﺘﺒﺮ ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ ﺑﺎﻟـ‬

Normal Probability Plot of the Residuals (response is C1)

Residual

10

0

-10

-2

-1

0

Normal Score

١٣٩

1

2

( ‫وهﻮ ﻣﻘﺒﻮل ) ﻧﻮﻋﺎ‬ ‫إذا ﻳﻤﻜﻦ إﻋﺘﺒﺎر اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح ﻣﻨﺎﺳﺒﺎ‬ ‫ ﺗﻨﺒﺆ ﻟﻬﺎ‬95% ‫اﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻟﻠﺘﻨﺒﺆات ﻟﺨﻤﺴﺔ ﻗﻴﻢ ﻣﺴﺘﻘﺒﻠﻴﺔ وﻓﺘﺮات‬

MTB > TSPlot C14 C15 C16; SUBC> Index; SUBC> TDisplay 11; SUBC> Symbol; SUBC> Connect; SUBC> overlay.

Time Series Plot for C1 (with forecasts and their 95% confidence limits)

36

C1

26

16

6

5

10

15

20

25

30

35

Time

:‫ﻣﺜﺎل ﺁﺧﺮ‬ ‫اﻟﻤﺸﺎهﺪات اﻟﺘﺎﻟﻴﺔ ﻟﻈﺎهﺮة ﻋﺸﻮاﺋﻴﺔ ﻣﺴﺠﻠﺔ ﻋﻠﻲ ﺷﻜﻞ ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ‬ 10.38 11.68 9.10 10.01 9.51 9.48 7.23 10.85 6.89

11.86 11.17 9.09 9.37 9.24 7.38 8.42 10.41 5.96

10.97 10.53 9.35 8.69 8.66 6.90 9.61 9.96 6.80

10.80 10.01 8.82 8.19 8.86 6.94 9.05 9.61 7.68

9.79 9.91 9.32 8.67 8.05 6.24 9.26 8.76 8.38

10.39 9.14 9.01 9.55 7.79 6.84 9.22 8.18 8.52

10.42 9.16 9.00 8.92 6.75 6.85 9.38 7.21 9.74

10.82 9.55 9.80 8.09 6.75 6.90 9.10 7.13 9.31

11.40 9.67 9.83 9.37 7.82 7.79 7.95 9.10 9.89

11.32 11.44 8.44 8.24 9.72 9.89 10.13 10.14 8.64 10.58 8.18 7.51 8.12 9.75 8.25 7.91 9.96

MTB > TSPlot C10; SUBC> Index; SUBC> TDisplay 11; SUBC> Symbol; SUBC> Connect.

12 11

C 1 0

10 9 8

١٤٠

7 6 Index

10

20

30

40

50

60

70

80

90

‫ﻧﻔﺤﺺ دوال اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬

Autocorrelation

Autocorrelation Function for C10 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

2

Lag Corr 1 2 3 4 5 6 7

0.83 0.61 0.46 0.37 0.33 0.28 0.26

T

12

LBQ

8.24 69.92 3.91 107.90 2.56 129.56 1.95 143.87 1.65 155.04 1.40 163.68 1.28 171.23

Lag Corr 8 9 10 11 12 13 14

0.26 0.26 0.18 0.09 0.04 0.03 0.04

T

LBQ

22

Lag Corr

1.26 178.83 1.21 186.14 0.84 189.86 0.43 190.87 0.20 191.09 0.13 191.19 0.19 191.39

T

LBQ

15 0.05 0.21 191.63 16 0.04 0.16 191.78 17 0.00 0.02 191.78 18 -0.03 -0.15 191.91 19 -0.05 -0.24 192.26 20 -0.05 -0.24 192.60 21 0.01 0.07 192.63

Lag Corr

T

LBQ

22 0.10 0.47 194.02 23 0.18 0.81 198.12 24 0.20 0.89 203.24

Partial Autocorrelation

Partial Autocorrelation Function for C10 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

2

12

Lag PAC

T

Lag PAC

T

1 0.83 2 -0.27 3 0.13 4 0.03 5 0.06 6 -0.02 7 0.09

8.24 -2.64 1.29 0.34 0.61 -0.21 0.91

8 0.05 9 0.00 10 -0.20 11 0.02 12 0.01 13 0.01 14 0.03

0.45 0.03 -1.98 0.19 0.09 0.12 0.34

22

Lag PAC 15 16 17 18 19 20 21

-0.01 -0.03 -0.07 -0.03 0.06 0.02 0.21

T

Lag PAC

T

-0.15 -0.25 -0.73 -0.26 0.60 0.20 2.03

22 0.05 23 0.06 24 -0.07

0.51 0.59 -0.65

AR ( 2 ) ‫ﻧﻼﺣﻆ ان اﻷﻧﻤﺎط ﺗﻘﺘﺮح ﻧﻤﻮذج إﻧﺤﺪار ذاﺗﻲ ﻣﻦ اﻟﺪرﺟﺔ اﻟﺜﺎﻧﻴﺔ‬ ‫ﻧﻄﺒﻖ اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح‬ MTB > Name c17 = 'RESI1' MTB > ARIMA 2 0 0 C10 'RESI1'; SUBC> Constant;

١٤١

SUBC> SUBC> SUBC> SUBC> SUBC>

Forecast 5 c14 c15 c16; GSeries; GACF; GPACF; GNormalplot.

ARIMA Model ARIMA model for C10 Estimates at each iteration Iteration SSE Parameters 0 126.398 0.100 0.100 1 103.515 0.250 0.043 2 84.535 0.400 -0.014 3 69.407 0.550 -0.071 4 58.132 0.700 -0.128 5 50.724 0.850 -0.184 6 47.212 1.000 -0.239 7 46.918 1.053 -0.256 8 46.916 1.054 -0.255 9 46.916 1.054 -0.255 10 46.916 1.054 -0.255 Relative change in each estimate less than Final Estimates of Parameters Type Coef StDev AR 1 1.0542 0.0992 AR 2 -0.2547 0.0993 Constant 1.81360 0.07092 Mean 9.0480 0.3538

7.283 6.434 5.586 4.738 3.887 3.030 2.163 1.838 1.816 1.814 1.814 0.0010

T 10.63 -2.56 25.57

Number of observations: 98 Residuals: SS = 46.7518 (backforecasts excluded) MS = 0.4921 DF = 95 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 48 Chi-Square 7.2(DF=10) 13.7(DF=22) 21.3(DF=34) 28.8(DF=46) Forecasts from period 98 Period 99 100 101 102 103

Forecast 9.7950 9.6033 9.4432 9.3232 9.2375

95 Percent Limits Lower Upper 8.4198 11.1703 7.6050 11.6016 7.1234 11.7629 6.8446 11.8018 6.6825 11.7925

Actual

‫اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح هﻮ‬ zt = 1.8136 + 1.0542 zt −1 − 0.2547 zt −2 + at , at WN ( 0,0.4921) ∀t ‫ ﻟﻬﺎ هﻲ‬t ‫ﻣﻘﺪرات اﻟﻤﻌﺎﻟﻢ وإﻧﺤﺮاﻓﺎﺗﻬﺎ اﻟﻤﻌﻴﺎرﻳﺔ وﻗﻴﻢ‬ φˆ1 = 1.0542 se φˆ1 = 0.0992 t = 10.63

( ) φˆ = −0.2547 se (φˆ ) = 0.0993 t = −2.56 δˆ = 1.8136 se (δˆ ) = 0.07092 t = 25.57 2

2

σˆ 2 = 0.4921 df = 95

α = 0.05 ‫ﻧﻼﺣﻆ ان ﺟﻤﻴﻊ اﻟﻤﻘﺪرات ﻣﻌﻨﻮﻳﺔ ﻋﻨﺪ‬ ١٤٢

‫ﻟﻜﻲ ﻧﻘﺒﻞ ﺑﻬﺬا اﻟﻨﻤﻮذج ﻋﻠﻲ اﻧﻪ ﻣﻨﺎﺳﺐ ﻟﻠﺘﻨﺒﺆ ﻧﺠﺮي إﺧﺘﺒﺎرات ﻋﻠﻲ اﻟﺒﻮاﻗﻲ‬ MTB > TTest 0.0 'RESI1'; SUBC> Alternative 0.

T-Test of the Mean Test of mu = 0.0000 vs mu not = 0.0000 Variable RESI1

N 98

Mean StDev -0.0082 0.6942

SE Mean 0.0701

T -0.12

P 0.91

‫ هﻲ إﺣﺘﻤﺎل ان اﻟﻔﺮﺿﻴﺔ اﻟﺼﻔﺮﻳﺔ‬P-Value ‫واﺿﺢ ﺟﺪا ان اﻷﺧﺘﺒﺎر ﻏﻴﺮ ﻣﻌﻨﻮي ) ﻣﻼﺣﻈﺔ اﻟـ‬ ( ‫ﺻﺤﻴﺤﺔ‬ ‫ﻧﺨﺘﺒﺮ ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬ MTB > Runs 'RESI1'.

Runs Test RESI1 K =

-0.0082

The observed number of runs = 47 The expected number of runs = 49.9184 47 Observations above K 51 below The test is significant at 0.5529 Cannot reject at alpha = 0.05

‫أي ﻻﻳﻤﻜﻨﻨﺎ رﻓﺾ ﻓﺮﺿﻴﺔ ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬ ‫ﻧﻔﺤﺺ اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ‬

ACF of Residuals for C10 (with 95% confidence limits for the autocorrelations) 1.0 0.8

Autocorrelation

0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

2

4

6

8

10

12

14

Lag

١٤٣

16

18

20

22

24

PACF of Residuals for C10 (with 95% confidence limits for the partial autocorrelations) 1.0

Partial Autocorrelation

0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

2

4

6

8

10

12

14

16

18

20

22

24

Lag

‫واﺿﺢ ﺟﺪا اﻧﻤﺎط اﻟﻀﺠﺔ اﻟﺒﻴﻀﺎء‬ ‫ﻳﺒﻘﻲ ﻓﺤﺺ ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬ Normal Probability Plot of the Residuals (response is C10) 2

Residual

1

0

-1

-2 -3

( at

-2

-1

0

1

Normal Score

2

3

IIDN ( 0,0.4921) ‫وﻧﺴﺘﻄﻴﻊ ان ﻧﻘﻮل ان اﻟﺒﻮاﻗﻲ ﻟﻬﺎ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ ) أي‬ ‫ ﺗﻨﺒﺆ ﻟﻬﺎ‬95% ‫اﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻟﻠﺘﻨﺒﺆات ﻟﺨﻤﺴﺔ ﻗﻴﻢ ﻣﺴﺘﻘﺒﻠﻴﺔ وﻓﺘﺮات‬

MTB > TSPlot C14 C15 C16; SUBC> Index; SUBC> TDisplay 11; SUBC> Symbol; SUBC> Connect; SUBC> overlay.

Time Series Plot for C10 (with forecasts and their 95% confidence limits) 12 11

C10

10 9 8 7 6

١٤٤

‫ﺗﻤﺮﻳﻦ‪ :‬ﻃﺒﻖ ﻋﻠﻲ اﻟﻤﺸﺎهﺪات اﻟﺴﺎﺑﻘﺔ ﻧﻤﻮذج )‪ AR (1‬وﻗﺎرن ﺑﻴﻦ اﻟﻨﺘﺎﺋﺞ‬

‫‪١٤٥‬‬

١٤٦

‫اﻟﻔﺼﻞ اﻟﺜﺎﻣﻦ‬ ‫ﻣﺜﺎل ﻋﻠﻰ ﺗﺤﻠﻴﻞ اﻟﺒﻮاﻗﻲ وﻣﻌﺎﻳﻴﺮ إﺧﺘﻴﺎر اﻟﻨﻤﻮذج اﻟﻤﻨﺎﺳﺐ‬ ‫‪Example on Residual Analysis and Model Selection‬‬ ‫‪: Criteria‬‬ ‫ﻟﻘﺪ ﻋﺮﻓﻨﺎ ﺳﺎﺑﻘﺎ اﻟﺒﻮاﻗﻲ ﻋﻠﻲ اﻧﻬﺎ اﻟﻘﻴﻢ اﻟﻤﺸﺎهﺪة ﻧﺎﻗﺺ اﻟﻘﻴﻢ اﻟﻤﻄﺒﻘﺔ ﻓﻤﻦ ﻣﺸﺎهﺪات ﻣﻌﻄﺎة‬ ‫‪ z1 , z2 ,..., zn‬وﻧﻤﻮذج ﻣﻄﺒﻖ ﻳﻨﺘﺞ ﻟﺪﻳﻨﺎ ﻗﻴﻢ ﻣﻄﺒﻘﺔ ‪ zˆ1 , zˆ2 ,..., zˆn‬وﺗﻜﺘﺐ اﻟﺒﻮاﻗﻲ‪:‬‬ ‫‪ei = zi − zˆi , i = 1, 2,..., n‬‬ ‫واﻟﺒﻮاﻗﻲ هﻲ ﻣﻘﺪرات اﻷﺧﻄﺎء ﻓﻲ اﻟﻨﻤﻮذج أي ‪ aˆi = ei , i = 1, 2,..., n‬وﻟﻬﺬا ﻳﺠﺐ ان ﺗﺤﻘﻖ‬ ‫اﻟﺸﺮوط اﻟﻤﻔﺮوﺿﺔ ﻋﻠﻲ اﻷﺧﻄﺎء ﻓﻲ هﺬا اﻟﻨﻤﻮذج واﻟﺘﻲ ﻣﻨﻬﺎ‪:‬‬ ‫‪ -١‬ﻣﺘﻮﺳﻂ اﻷﺧﻄﺎء ﻳﺴﺎوي اﻟﺼﻔﺮ‬ ‫‪ -٢‬اﻷﺧﻄﺎء ﻋﺸﻮاﺋﻴﺔ و ﻏﻴﺮ ﻣﺘﺮاﺑﻄﺔ أو ﻣﺴﺘﻘﻠﺔ ) وﻓﻲ آﺜﻴﺮ ﻣﻦ اﻟﻨﻤﺎذج ﻧﻔﺘﺮض ان اﻷﺧﻄﺎء‬ ‫ﻟﻬﺎ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ ﻣﺴﺘﻘﻞ وﻣﺘﻄﺎﺑﻖ ﺑﻤﺘﻮﺳﻂ ﺻﻔﺮي وﺗﺒﺎﻳﻦ ‪ σ 2‬أي ) ‪( at IIDN ( 0, σ 2‬‬ ‫ﻟﻬﺬا ﻓﺈﻧﻨﺎ ﻧﺠﺮي ﺗﺤﻠﻴﻼ وهﻮ ﻣﺠﻤﻮﻋﺔ ﻣﻦ اﻹﺧﺘﺒﺎرات ﻋﻠﻲ اﻟﺒﻮاﻗﻲ ﻟﻨﺮي ﻓﻴﻤﺎ إذا آﺎﻧﺖ ﺗﺤﻘﻖ هﺬﻩ‬ ‫اﻟﺸﺮوط وﻓﻲ هﺬﻩ اﻟﺤﺎﻟﺔ ﻧﻌﺘﺒﺮ اﻟﻨﻤﻮذج اﻟﻤﻄﺒﻖ ﻣﻘﺒﻮﻻ أﻣﺎ إذا ﻓﺸﻞ اﺣﺪ هﺬﻩ اﻹﺧﺘﺒﺎرات ﻓﻴﺠﺐ‬ ‫ﻋﻠﻴﻨﺎ إﻋﺎدة اﻟﻨﻈﺮ وإﻗﺘﺮاح ﻧﻤﻮذج ﺁﺧﺮ‬ ‫أوﻻ‪ :‬إﺧﺘﺒﺎر اﻟﻤﺘﻮﺳﻂ‬ ‫‪H 0 : E ( at ) = 0‬‬ ‫‪H 1 : E ( at ) ≠ 0‬‬

‫وهﻮ إﺧﺘﺒﺎر ﺑﺬﻳﻠﻴﻦ وﻧﺴﺘﺨﺪم ﻓﻴﺔ اﻹﺣﺼﺎﺋﺔ‬

‫‪e‬‬

‫) ‪se ( e‬‬

‫= ‪ u‬واﻟﺘﻲ ﻟﻬﺎ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ ﻗﻴﺎﺳﻲ ﻓﻌﻨﺪ‬

‫ﻣﺴﺘﻮى ﻣﻌﻨﻮﻳﺔ ‪ α = 0.05‬ﻧﻌﺘﺒﺮ ان ‪ E ( at ) = 0‬إذا آﺎﻧﺖ ‪ ) u < 1.96‬هﺬا ﻋﻠﻲ إﻋﺘﺒﺎر ان‬ ‫ﺣﺠﻢ اﻟﻌﻴﻨﺔ اآﺒﺮ ﻣﻦ ‪ ٣٠‬وﺣﺪة وهﺬا داﺋﻤﺎ ﻣﺘﺤﻘﻖ ﻟﻠﻤﺘﺴﻠﺴﻼت اﻟﺰﻣﻨﻴﺔ اﻟﺘﻲ ﻧﺪرﺳﻬﺎ (‬

‫ﻣﺜﺎل‪:‬‬ ‫ﺳﻮف ﻧﻌﻮد اﻟﻲ ﻣﺜﺎل ﺗﻄﺒﻴﻖ ﻣﺘﻮﺳﻂ ﻣﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ اﻟﺜﺎﻟﺜﺔ ﻋﻠﻲ اﻟﻤﺘﻐﻴﺮ ‪Metals‬‬ ‫‪MTB > RETR 'E:\Mtbwin\DATA\EMPLOY.MTW'.‬‬ ‫‪Retrieving worksheet from file: E:\Mtbwin\DATA\EMPLOY.MTW‬‬ ‫‪Worksheet was saved on 6/ 5/1996‬‬ ‫;'‪MTB > TSPlot 'Metals‬‬ ‫>‪SUBC‬‬ ‫;‪Index‬‬ ‫>‪SUBC‬‬ ‫;‪TDisplay 11‬‬ ‫>‪SUBC‬‬ ‫;‪Symbol‬‬ ‫>‪SUBC‬‬ ‫‪Connect.‬‬

‫‪50‬‬

‫‪M‬‬ ‫‪e‬‬ ‫‪ta‬‬ ‫‪ls‬‬

‫‪45‬‬

‫‪40‬‬ ‫‪60‬‬

‫‪50‬‬

‫‪40‬‬

‫‪30‬‬

‫‪20‬‬

‫‪10‬‬

‫‪Index‬‬

‫'‪MTB > Name c4 = 'AVER1' c5 = 'FITS1' c6 = 'RESI1‬‬

‫‪١٤٧‬‬

MTB > %MA 'Metals' 3; SUBC> Averages 'AVER1'; SUBC> Fits 'FITS1'; SUBC> Residuals 'RESI1'. Executing from file: E:\MTBWIN\MACROS\MA.MAC

Moving average Data Length NMissing

Metals 60.0000 0

Moving Average Length: 3 Accuracy Measures MAPE: 1.55036 MAD: 0.70292 MSD: 0.76433

Moving Average

Actual

Predicted

M etals

50

Actual Predicted

45 Moving Average Length:

40 0

10

20

30

40

50

3

MAPE:

1.55036

MAD:

0.70292

MSD:

0.76433

60

Time

‫ﻻﺣﻆ اﻧﻨﺎ ﺧﺰﻧﺎ اﻟﺒﻮاﻗﻲ ﻓﻲ اﻟﻌﻤﻮد اﻟﺴﺎدس واﻟﻘﻴﻢ اﻟﻤﻄﺒﻘﺔ ﻓﻲ اﻟﻌﻤﻮد اﻟﺨﺎﻣﺲ‬ MTB > print c3 c6 c5

Data Display

Row

Metals

RESI1

FITS1

1 2 3 4 5 6 7 8 9

44.2 44.3 44.4 43.4 42.8 44.3 44.4 44.8 44.4

* * * -0.90000 -1.23333 0.76667 0.90000 0.96667 -0.10000

* * * 44.3000 44.0333 43.5333 43.5000 43.8333 44.5000

١٤٨

10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

43.1 42.6 42.4 42.2 41.8 40.1 42.0 42.4 43.1 42.4 43.1 43.2 42.8 43.0 42.8 42.5 42.6 42.3 42.9 43.6 44.7 44.5 45.0 44.8 44.9 45.2 45.2 45.0 45.5 46.2 46.8 47.5 48.3 48.3 49.1 48.9 49.4 50.0 50.0 49.6 49.9 49.6 50.7 50.7 50.9 50.5 51.2 50.7 50.3 49.2 48.1

-1.43333 -1.50000 -0.96667 -0.50000 -0.60000 -2.03333 0.63333 1.10000 1.60000 -0.10000 0.46667 0.33333 -0.10000 -0.03333 -0.20000 -0.36667 -0.16667 -0.33333 0.43333 1.00000 1.76667 0.76667 0.73333 0.06667 0.13333 0.30000 0.23333 -0.10000 0.36667 0.96667 1.23333 1.33333 1.46667 0.76667 1.06667 0.33333 0.63333 0.86667 0.56667 -0.20000 0.03333 -0.23333 1.00000 0.63333 0.56667 -0.26667 0.50000 -0.16667 -0.50000 -1.53333 -1.96667

44.5333 44.1000 43.3667 42.7000 42.4000 42.1333 41.3667 41.3000 41.5000 42.5000 42.6333 42.8667 42.9000 43.0333 43.0000 42.8667 42.7667 42.6333 42.4667 42.6000 42.9333 43.7333 44.2667 44.7333 44.7667 44.9000 44.9667 45.1000 45.1333 45.2333 45.5667 46.1667 46.8333 47.5333 48.0333 48.5667 48.7667 49.1333 49.4333 49.8000 49.8667 49.8333 49.7000 50.0667 50.3333 50.7667 50.7000 50.8667 50.8000 50.7333 50.0667

‫اﻵن ﻧﺨﺘﺒﺮ ﻣﺘﻮﺳﻂ اﻟﺒﻮاﻗﻲ‬ MTB > TTest 0.0 'RESI1'; SUBC> Alternative 0.

T-Test of the Mean

١٤٩

Test of mu = 0.000 vs mu not = 0.000 Variable RESI1

N 57

Mean 0.158

StDev 0.868

SE Mean 0.115

T 1.37

P 0.17

‫ ﻳﺴﺘﺨﺪم ﺑﺮﻧﺎﻣﺞ ﻋﺎم ﻋﻨﺪﻣﺎ ﻳﻜﻮن اﻹﻧﺤﺮاف اﻟﻤﻌﻴﺎري ) او اﻟﺘﺒﺎﻳﻦ ( ﻏﻴﺮ‬Minitab ‫ ﻓﻲ‬:‫ﻣﻼﺣﻈﺔ‬ ‫ أي‬١٫٩٦ ‫ وهﻲ اﻗﻞ ﻣﻦ‬T=١٫٣٧ ‫ ﻻﺣﻆ ان ﻗﻴﻤﺔ اﻹﺣﺼﺎﺋﺔ هﻲ‬. Ttest ‫ﻣﻌﺮوف وﻳﻄﻠﻖ ﻋﻠﻴﻪ‬ ‫ﻻﻧﺮﻓﺾ اﻟﻔﺮﺿﻴﺔ اﻟﺼﻔﺮﻳﺔ‬ ‫ ﺣﻮل اﻟﻤﺘﻮﺳﻂ وﺣﻮل اﻟﺼﻔﺮ‬Runs test ‫ ﻧﺨﺘﺒﺮ ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ ﺑﻮاﺳﻄﺔ إﺧﺘﺒﺎر اﻟﺠﺮي‬:‫ﺛﺎﻧﻴﺎ‬ :‫ﺗﺎﺑﻊ اﻟﻤﺜﺎل‬ MTB > Runs 'RESI1'.

Runs Test

RESI1 K =

0.1579

The observed number of runs = 17 The expected number of runs = 29.4211 30 Observations above K 27 below The test is significant at 0.0009 MTB > Runs 0 'RESI1'.

Runs Test

RESI1 K =

0.0000

The observed number of runs = 17 The expected number of runs = 28.7895 33 Observations above K 24 below The test is significant at 0.0013

‫ﻧﻼﺣﻆ اﻧﻪ ﻓﻲ آﻠﺘﺎ اﻟﺤﺎﻟﺘﻴﻦ ﻻﻧﺮﻓﺾ ﻋﺸﻮاﺋﻴﺔ اﻟﺒﻮاﻗﻲ‬ Autocorrelation test ‫ ﻧﺨﺘﺒﺮ ﺗﺮاﺑﻂ أو إﺳﺘﻘﻼل اﻟﺒﻮاﻗﻲ ﺑﻮاﺳﻄﺔ إﺧﺘﺒﺎر اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ‬:‫ﺛﺎﻟﺜﺎ‬ :‫ﺗﺎﺑﻊ اﻟﻤﺜﺎل‬ MTB > %ACF 'RESI1'. Executing from file: E:\MTBWIN\MACROS\ACF.MAC

١٥٠

Autocorrelation

Autocorrelation Function for RESI1 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

4

9

14

Lag

Corr

T

LBQ

Lag

Corr

T

LBQ

1 2 3 4 5 6 7

0.56 0.24 -0.01 0.04 0.02 -0.07 -0.14

4.24 1.39 -0.06 0.21 0.11 -0.38 -0.81

18.93 22.31 22.32 22.40 22.42 22.72 24.07

8 9 10 11 12 13 14

-0.04 0.12 0.29 0.24 0.20 0.05 0.03

-0.22 0.68 1.64 1.30 1.05 0.25 0.17

24.17 25.21 31.36 35.67 38.71 38.90 38.99

‫ أي‬ρ1 = 0 ‫ أي اﻧﻨﺎ ﻧﺮﻓﺾ ان‬٤٫٢٤ ‫ ﻟﻠﺘﺮاﺑﻂ اﻟﺬاﺗﻲ ﻋﻨﺪ اﻟﺘﺨﻠﻒ اﻷول ﺗﺴﺎوي‬T ‫ﻻﺣﻆ ان اﻟـ‬ ‫ﻳﻮﺟﺪ ﺗﺮاﺑﻂ ﺑﻴﻦ اﻟﺒﻮاﻗﻲ ﻣﻦ اﻟﺪرﺟﺔ اﻻوﻟﻰ ﻓﻲ اﻹﺧﺘﺒﺎر‬ H 0 : ρ1 = 0

H1 : ρ1 ± 0 r1 = 4.24 ‫ﺣﻴﺚ اﻹﺣﺼﺎﺋﺔ هﻲ‬ se ( r1 ) ‫ ﻧﺨﺘﺒﺮ ﻓﻲ ﻣﺎ إذا آﺎﻧﺖ اﻟﺒﻮاﻗﻲ ﻣﻮزﻋﺔ ﻃﺒﻴﻌﻴﺎ‬:‫راﺑﻌﺎ‬ :‫ﺗﺎﺑﻊ اﻟﻤﺜﺎل‬ MTB > %NormPlot 'RESI1'; SUBC> Kstest. Executing from file: E:\MTBWIN\MACROS\NormPlot.MAC

Normal Probability Plot

.999

P ro b a b ility

.99 .95 .80 .50 .20 .05 .01 .001 -2

-1

0

1

RESI1 Average: 0.157895 StDev: 0.867525 N: 57

Kolmogorov-Smirnov Normality Test D+: 0.054 D-: 0.084 D : 0.084 Approximate P-Value > 0.15

‫ أي ﻻﻧﺮﻓﺾ ﻓﺮﺿﻴﺔ اﻟﺘﻮزﻳﻊ‬٠٫٠٥ ‫ وهﻲ اآﺒﺮ ﻣﻦ‬٠٫١٥ ‫ اﻟﻨﺎﺗﺠﺔ ﺗﺴﺎوي‬P-Value ‫ﻻﺣﻆ اﻟـ‬ ‫ واﻟﺬي ﻳﺒﻴﻦ ﻣﺪي‬Q-Q Plot ‫ هﻨﺎك اﻳﻀﺎ إﺧﺘﺒﺎر ﺁﺧﺮ ﻟﻠﻄﺒﻴﻌﻴﺔ هﻮ اﻟـ‬α = 0.05 ‫اﻟﻄﺒﻴﻌﻲ ﻋﻨﺪ‬ ‫ﺗﻄﺎﺑﻖ ﻣﺸﺎهﺪات ﻣﺎ ﻣﻊ ﺗﻮزﻳﻊ ﻣﻌﻴﻦ‬ MTB > %Qqplot 'RESI1'; SUBC> Table; SUBC> Conf 95; SUBC> Ci. Executing from file: E:\MTBWIN\MACROS\Qqplot.MAC

١٥١

‫‪Distribution Function Analysis‬‬

‫‪Normal Dist. Parameter Estimates‬‬ ‫‪Data‬‬

‫‪: RESI1‬‬

‫‪Mean:‬‬ ‫‪StDev:‬‬

‫‪0.157895‬‬ ‫‪0.867525‬‬

‫‪Normal Probability Plot for RESI1‬‬

‫‪99‬‬

‫‪0.157895‬‬

‫‪Mean:‬‬

‫‪0.867525‬‬

‫‪StDev:‬‬ ‫‪95‬‬ ‫‪90‬‬ ‫‪80‬‬

‫‪60‬‬ ‫‪50‬‬ ‫‪40‬‬

‫‪Percent‬‬

‫‪70‬‬

‫‪30‬‬ ‫‪20‬‬ ‫‪10‬‬ ‫‪5‬‬

‫‪1‬‬

‫‪2‬‬

‫‪0‬‬

‫‪1‬‬

‫‪-1‬‬

‫‪-2‬‬

‫‪Data‬‬

‫ﻻﺣﻆ ان ﻓﻲ آﻠﺘﺎ اﻟﺤﺎﻟﺘﻴﻦ ﻓﺈﻧﻨﺎ ﻻﻧﺮﻓﺾ ان اﻟﺒﻮاﻗﻲ ﻟﻬﺎ ﺗﻮزﻳﻊ ﻃﺒﻴﻌﻲ‬ ‫ﻣﻼﺣﻈﺔ اﺧﻴﺮة‪ :‬ﻳﺒﺪو ان اﻟﺒﻮاﻗﻲ ﺗﺤﻘﻖ ﻣﻌﻈﻢ اﻟﺸﺮوط ﻓﻴﻤﺎ ﻋﺪي اﻟﺘﺮاﺑﻂ اﻟﺬي ﻳﻮﺟﺪ ﺑﻴﻦ اﻟﻘﻴﻢ‬ ‫اﻟﻤﺘﺘﺎﻟﻴﺔ وهﺬا ﻳﺠﻌﻠﻨﺎ ﻧﺮﻓﺾ ﺟﻮدة اﻟﺘﻄﺒﻴﻖ ﻟﻄﺮﻳﻘﺔ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ اﻟﺜﺎﻟﺜﺔ ﺣﻴﺚ‬ ‫ادي اﻟﻲ ﺑﻮاﻗﻲ ﻣﺘﺮاﺑﻄﺔ‪.‬‬

‫‪١٥٢‬‬

‫اﻟﻔﺼﻞ اﻟﺘﺎﺳﻊ‬ ‫ﺗﺤﻠﻴﻞ او ﺗﻔﻜﻴﻚ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻲ ﻣﺮآﺒﺎت ‪:Decomposition Method‬‬ ‫ﻳﻨﻈﺮ إﻟﻰ اﻟﻤﺘﻠﺴﻠﺴﺔ اﻟﺰﻣﻨﻴﺔ ﻋﻠﻰ اﻧﻬﺎ ﻣﻜﻮﻧﺔ ﻣﻦ ﻋﺪة ﻣﺮآﺒﺎت أو اﺟﺰاء ﻣﺘﺤﺪة ﻣﻊ ﺑﻌﻀﻬﺎ ﻟﺘﻜﻮﻳﻦ‬ ‫هﺬﻩ اﻟﻤﺘﺴﻠﺴﻠﺔ‪،‬‬ ‫ﻟﻨﻔﺘﺮض ان ﻟﺪﻳﻨﺎ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ اﻟﻤﺸﺎهﺪﻩ ‪ . z1 , z2 ,..., zn‬ﻟﻘﺪ وﺟﺪ ﺑﺎﻟﺘﺠﺮﺑﺔ أﻧﻪ ﻳﻤﻜﻦ ﻧﻤﺬﺟﺘﻬﺎ‬ ‫ﻋﻠﻰ اﻟﺸﻜﻞ‬ ‫‪zt = Tt + St + Ct + Et , t = 1, 2,..., n‬‬

‫ﺣﻴﺚ ‪ Zt‬اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ اﻟﻤﺸﺎهﺪة و ‪ Tt‬ﻣﺮآﺒﺔ اﻹﻧﺠﺮاف وهﻲ اﻟﺘﻲ ﺗﻨﻤﺬج اﻹﺗﺠﺎﻩ اﻟﻌﺎم اﻟﺬي‬ ‫ﺗﻨﺤﻲ أو ﺗﻨﺠﺮف اﻟﻴﻪ اﻟﻤﺘﺴﻠﺴﻠﺔ و ‪ St‬ﻣﺮآﺒﺔ ﻣﻮﺳﻤﻴﺔ وﺗﻨﻤﺬج اﻟﺘﺄﺛﻴﺮ اﻟﻤﻮﺳﻤﻲ )إذا وﺟﺪ( وهﻮ‬ ‫اﻟﺘﻐﻴﺮ اﻟﺬي ﻳﺤﺪث ﻟﻠﻤﺘﺴﻠﺴﻠﺔ ﻧﺘﻴﺠﺔ اﻟﺘﺄﺛﻴﺮات اﻟﻤﻮﺳﻤﻴﺔ ﻣﺜﻞ اﻟﺸﻬﺮﻳﺔ واﻟﺴﻨﻮﻳﺔ و ‪ Ct‬ﻣﺮآﺒﺔ‬ ‫دورﻳﺔ )إذا وﺟﺪت( و ﺗﻨﻤﺬج ﻣﻨﺤﻰ أو إﺗﺠﺎة ﻳﺘﻜﺮر ﺑﻌﺪ ﻓﺘﺮات زﻣﻨﻴﺔ ﻃﻮﻳﻠﺔ ﻏﻴﺮ ﻣﻮﺳﻤﻴﺔ و ‪Et‬‬ ‫ﻣﺮآﺒﺔ اﻟﺨﻄﺄ وﺗﺸﻤﻞ ﺟﻤﻴﻊ اﻟﻌﻮاﻣﻞ اﻻﺧﺮى اﻟﺘﻲ ﺗﺆﺛﺮ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ واﻟﺘﻲ ﻻﻳﻤﻜﻦ ﻧﻤﺬﺟﺘﻬﺎ‬ ‫ﺿﻤﻨﻴﺎ أو اﻟﺘﻲ ﻻﻳﻤﻜﻦ ﻣﺸﺎهﺪﺗﻬﺎ او ﻗﻴﺎﺳﻬﺎ‪ .‬اﻟﻨﻤﻮذج اﻟﺴﺎﺑﻖ ﻳﺴﻤﻲ ﺑﺎﻟﻨﻤﻮذج اﻹﺿﺎﻓﻲ ‪Additive‬‬ ‫‪ Model‬وذﻟﻚ ﻷن آﻞ اﻟﻤﺮآﺒﺎت ﺗﺪﺧﻞ ﺑﺸﻜﻞ إﺿﺎﻓﻲ ﻓﻲ اﻟﻨﻤﻮذج‪ .‬هﻨﺎك أﺷﻜﺎل اﺧﺮى ﻣﺜﻞ‬ ‫‪zt = Tt St + Ct + Et , t = 1, 2,..., n‬‬ ‫‪zt = Tt St Ct + Et , t = 1, 2,..., n‬‬

‫واﻟﺘﻲ ﺗﺴﻤﻰ ﺑﺎﻟﻨﻤﺎذج اﻟﺘﻀﺎﻋﻔﻴﺔ ‪. Multiplicative Models‬‬ ‫ﻓﻲ هﺬا اﻟﻤﺴﺘﻮى ﺳﻮف ﻧﻬﻤﻞ اﻟﻤﺮآﺒﺔ اﻟﺪورﻳﺔ ‪ Ct‬وذﻟﻚ ﻷن اﻟﻤﺮآﺒﺔ اﻟﺪورﻳﺔ ﻧﺎدرا ﻣﺎﺗﻜﻮن‬ ‫ﻣﻮﺟﻮدة ﻓﻲ اﻟﻤﺘﺴﻠﺴﻼت اﻟﻘﺼﻴﺮة أو اﻟﻄﻮﻳﻠﺔ ﻧﺴﺒﻴﺎ ﻷﻧﻬﺎ ﺗﺤﺘﺎج اﻟﻰ ﻣﺸﺎهﺪات ﻃﻮﻳﻠﺔ ﺟﺪا ﻋﻠﻰ‬ ‫ﻣﺪي ﻋﺪد آﺒﻴﺮ ﻣﻦ اﻟﻌﻘﻮد‪.‬‬ ‫وﻧﻜﺘﻔﻲ ﺑﺎﻟﻨﻤﺎذج ﻋﻠﻲ اﻟﺸﻜﻞ‬ ‫‪zt = Tt + St + Et , t = 1,2,..., n‬‬

‫‪zt = Tt St + Et , t = 1,2,..., n‬‬ ‫أﻧﻈﺮ آﺘﺎب ‪ FORECASTING: METHODS AND APPLICATIONS‬ﻟﻠﻤﺆﻟﻔﻴﻦ‬ ‫‪ MAKRIDAKIS/ WHEELWRIGHT/ McGEE‬ص ‪١٤١-١٣١‬‬ ‫ﺳﻮف ﻧﺴﺘﻌﺮض ﻓﻲ اﻟﻤﺜﺎل اﻟﺘﺎﻟﻲ ﻃﺮق اﻟﺘﺤﻠﻴﻞ اﻹﺿﺎﻓﻴﺔ واﻟﺘﻀﺎﻋﻔﻴﺔ ﺑﺪون اﻟﻤﺮآﺒﺔ اﻟﺪورﻳﺔ أي‬ ‫اﻟﻨﻤﺎذج‬ ‫‪zt = Tt + St + Et , t = 1,2,..., n‬‬

‫‪zt = Tt St + Et , t = 1,2,..., n‬‬ ‫اﻟﺒﻴﺎﻧﺎت اﻟﺘﺎﻟﻴﺔ هﻲ اﻟﻄﻠﺐ ﻋﻠﻲ اﻟﺒﻨﺰﻳﻦ ﺑﻤﻼﻳﻴﻦ اﻟﻠﺘﺮات ﻓﻲ ﻣﺪﻳﻨﺔ اوﻧﺘﺎرﻳﻮ ﺑﻜﻨﺪة ﻣﻦ ﺳﻨﺔ ‪١٩٦٠‬‬ ‫وﺣﺘﻰ ﺳﻨﺔ ‪١٩٧٥‬‬ ‫‪GasDemand‬‬ ‫‪MONTHLY GASOLINE DEMAND ONTARIO GALLON MILLIONS 1960-1975‬‬ ‫‪87695 86890 96442 98133 113615 123924 128924 134775 117357 114626‬‬ ‫‪107677 108087 92188 88591 98683 99207 125485 124677 132543 140735‬‬ ‫‪124008 121194 111634 111565 101007 94228 104255 106922 130621 125251‬‬ ‫‪140318 146174 122318 128770 117518 115492 108497 100482 106140 118581‬‬ ‫‪132371 132042 151938 150997 130931 137018 121271 123548 109894 106061‬‬ ‫‪112539 125745 136251 140892 158390 148314 144148 140138 124075 136485‬‬ ‫‪109895 109044 122499 124264 142296 150693 163331 165837 151731 142491‬‬ ‫‪140229 140463 116963 118049 137869 127392 154166 160227 165869 173522‬‬ ‫‪155828 153771 143963 143898 124046 121260 138870 129782 162312 167211‬‬ ‫‪172897 189689 166496 160754 155582 145936 139625 137361 138963 155301‬‬ ‫‪172026 165004 185861 190270 163903 174270 160272 165614 146182 137728‬‬ ‫‪148932 156751 177998 174559 198079 189073 175702 180097 155202 174508‬‬

‫‪١٥٣‬‬

154277 174176 198688 218099 227443 193522 199024 217775

144998 184416 190474 229001 233038 212870 191813 227621

159644 158167 194502 203200 234119 248565 195997

168646 156261 190755 212557 255133 221532 208684

166273 176353 166286 197095 216478 252642 244113

190176 175720 170699 193693 232868 255007 243108

205541 193939 181468 188992 221616 206826 255918

193657 201269 174241 175347 209893 233231 244642

182617 218960 210802 196265 194784 212678 237579

189614 209861 212262 203526 189756 217173 237579

‫أوﻻ ﻧﺮﺳﻢ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻓﻲ ﻣﺨﻄﻂ زﻣﻨﻲ‬ MTB > TSPlot 'GasDemand'; SUBC> Index; SUBC> TDisplay 11; SUBC> Symbol; SUBC> Connect.

G asD em and

250

200

150

100

100 150 ‫اﻟﻰ اﻻﻋﻠﻰ‬ ‫ﻣﻮﺳﻤﻴﺔ وﻣﻨﺠﺮﻓﺔ‬ ‫ﻧﻼﺣﻆ ان اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ‬ zt = Tt + St + Et , t = 1, 2,..., n :‫ ﺗﻄﺒﻴﻖ اﻟﻨﻤﻮذج اﻹﺿﺎﻓﻲ‬:‫اوﻻ‬ 50

Index

SUBC> SUBC> SUBC> SUBC>

MTB > %Decomp 'GasDemand' 12; Additive ; Forecasts 24; Title "Forecast of Gasoline Demand"; Start 1.

Time Series Decomposition Data Length NMissing

GasDeman 192.000 0

Trend Line Equation Yt = 96.4074 + 0.680579*t Seasonal Indices Period 1 2 3 4 5 6 7 8 9 10 11 12

Index -20.5625 -26.8125 -14.8958 -11.0625 9.89583 11.8958 22.7708 25.1875 5.64583 7.27083 -4.81250 -4.52083

١٥٤

Accuracy of Model MAPE: MAD: MSD:

3.6952 5.6622 52.7851

Forecasts Row

Period

Forecast

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216

207.197 201.627 214.225 218.738 240.377 243.058 254.614 257.711 238.850 241.155 229.753 230.725 215.364 209.794 222.391 226.905 248.544 251.225 262.780 265.878 247.017 249.322 237.919 238.892

Forecast of Gasoline Demand Seasonal Indices

Original Data, by Seasonal Period

30

250

20 10

200

0 150

-10 -20

100

-30 1 2 3 4 5 6 7 8 9 10 11 12

1 2 3 4 5 6 7 8 9 10 11 12

Percent Variation, by Seasonal Period

Residuals, by Seasonal Period 30 20

10

10 0 5

-10 -20

0

-30 1 2 3 4 5 6 7 8 9 10 11 12

١٥٥

1 2 3 4 5 6 7 8 9 10 11 12

(١) ‫ﺷﻜﻞ‬

‫ﺷﻜﻞ )‪(٢‬‬ ‫‪Forecast of Gasoline Demand‬‬ ‫‪Original Data‬‬

‫‪Detrended Data‬‬ ‫‪50‬‬ ‫‪40‬‬ ‫‪30‬‬ ‫‪20‬‬ ‫‪10‬‬ ‫‪0‬‬ ‫‪-10‬‬ ‫‪-20‬‬ ‫‪-30‬‬ ‫‪-40‬‬ ‫‪100‬‬

‫‪200‬‬

‫‪250‬‬ ‫‪200‬‬ ‫‪150‬‬ ‫‪100‬‬ ‫‪200‬‬

‫‪0‬‬

‫‪Seasonally Adj. and Detrended Data‬‬

‫‪0‬‬

‫‪100‬‬

‫‪Seasonally Adjusted Data‬‬ ‫‪250‬‬

‫‪30‬‬ ‫‪20‬‬ ‫‪10‬‬

‫‪200‬‬

‫‪-10‬‬

‫‪150‬‬

‫‪0‬‬ ‫‪-20‬‬ ‫‪-30‬‬ ‫‪100‬‬

‫‪200‬‬

‫ﺷﻜﻞ )‪(٣‬‬

‫‪100‬‬ ‫‪200‬‬

‫‪0‬‬

‫‪0‬‬

‫‪100‬‬

‫‪Forecast of Gasoline Demand‬‬

‫‪Actual‬‬ ‫‪Predicted‬‬

‫‪250‬‬

‫‪Actual‬‬ ‫‪Predicted‬‬ ‫‪Forecast‬‬

‫‪200‬‬

‫‪Forecast‬‬

‫‪GasDeman‬‬

‫‪150‬‬

‫‪100‬‬ ‫‪3.6952‬‬ ‫‪5.6622‬‬ ‫‪52.7851‬‬

‫‪MAPE:‬‬ ‫‪MAD:‬‬ ‫‪MSD:‬‬

‫‪200‬‬

‫‪100‬‬

‫‪0‬‬

‫‪Time‬‬

‫ﻣﻨﺎﻗﺸﺔ اﻟﻨﺘﺎﺋﺞ‪:‬‬ ‫ﺷﻜﻞ )‪ (١‬ﻳﻮﺿﺢ اﻟﻤﺆﺷﺮات اﻟﻤﻮﺳﻤﻴﺔ ‪ ، Seasonal Indices‬ﻓﺎﻟﺸﻜﻞ اﻷﻋﻠﻰ ﻣﻦ اﻟﻴﺴﺎر ﻳﺒﻴﻦ‬ ‫ﺗﺄﺛﺮاﻟﻄﻠﺐ ﻓﻲ اﻷﺷﻬﺮ اﻟﻤﺨﺘﻠﻔﺔ ﻣﻦ اﻟﺴﻨﺔ ﻓﻔﻲ اﻷﺷﻬﺮ ‪ ١١‬و ‪ ١٢‬و ‪ ١‬و ‪ ٢‬و ‪ ٣‬و ‪ ٤‬ﻳﺤﺪث ﻧﻘﺺ‬ ‫ﻓﻲ اﻟﻄﻠﺐ إذ ﻳﺘﻨﺎﻗﺺ ﺗﺪرﻳﺠﻴﺎ ﺣﺘﻰ ﻳﺼﻞ إﻟﻰ أﻗﻞ ﻣﻌﺪل ﻟﻪ ﻓﻲ اﻟﺸﻬﺮ ‪ ٢‬ﺛﻢ ﻳﺘﺰاﻳﺪ ﺣﺘﻰ ﻳﺼﺒﺢ‬ ‫ﻣﻮﺟﺒﺎ ﻓﻲ اﻟﺸﻬﺮ ‪ ٥‬وﻳﺘﺰاﻳﺪ ﺣﺘﻰ ﻳﺼﻞ اﻗﺼﻰ ﻗﻴﻤﺔ ﻣﻮﺟﺒﺔ ﻓﻲ اﻟﺸﻬﺮ ‪ ٨‬ﺛﻢ ﻳﻨﻘﺺ ﺑﺸﻜﻞ آﺒﻴﺮ‬ ‫ﺑﻌﺪﺋﺬ‪ .‬اﻟﺸﻜﻞ اﻷﻋﻠﻰ ﻣﻦ اﻟﻴﻤﻴﻦ ﻳﻌﻄﻲ رﺳﻢ اﻟﺼﻨﺪوق ‪ Box Plot‬ﻟﻠﻤﺸﺎهﺪات اﻷﺻﻠﻴﺔ ﻣﻮزﻋﺔ‬ ‫ﻋﻠﻰ اﻷﺷﻬﺮ وهﻮ ﻳﻮﺿﺢ ﺗﻮزﻳﻊ وإﻧﺘﺸﺎر اﻟﻤﺸﺎهﺪات ﻋﻠﻰ آﻞ ﺷﻬﺮ واﻟﻘﻴﻢ اﻟﺨﺎرﺟﺔ ‪. Out Liers‬‬ ‫اﻟﺸﻜﻞ اﻷﺳﻔﻞ ﻣﻦ اﻟﻴﺴﺎر ﻳﻌﻄﻲ اﻟﺘﻐﻴﺮ اﻟﻨﺴﺒﻲ اﻟﻤﺌﻮي ﻋﻠﻰ اﻟﻔﺘﺮات اﻟﻤﻮﺳﻤﻴﺔ )اﻷﺷﻬﺮ(‪ .‬اﻟﺸﻜﻞ‬ ‫اﻷﺳﻔﻞ اﻷﻳﻤﻦ ﻳﻌﻄﻲ رﺳﻢ اﻟﺼﻨﺪوق ﻟﻠﺒﻮاﻗﻲ أو اﻷﺧﻄﺎء ﻣﻮزﻋﺔ ﻋﻠﻰ اﻷﺷﻬﺮ‪.‬‬ ‫ﺷﻜﻞ )‪ (٢‬اﻟﺸﻜﻞ اﻷﻋﻠﻰ ﻣﻦ اﻟﻴﻤﻴﻦ ﻳﻌﻄﻲ اﻟﻤﺸﺎهﺪات اﻷﺻﻠﻴﺔ‪ ،‬اﻟﺸﻜﻞ اﻷﻋﻠﻰ ﻣﻦ اﻟﻴﺴﺎر ﻳﻌﻄﻲ‬ ‫اﻟﻤﺸﺎهﺪات ﺑﻌﺪ إزاﺣﺔ ﻣﺮآﺒﺔ اﻹﻧﺠﺮاف أي‬ ‫‪wt = zt − Tt , t = 1, 2,..., n‬‬ ‫‪=St + Et , t = 1, 2,..., n‬‬

‫‪١٥٦‬‬

‫اﻟﺸﻜﻞ اﻷﺳﻔﻞ ﻣﻦ اﻟﻴﺴﺎر ﻳﻌﻄﻲ اﻟﻤﺸﺎهﺪات اﻷﺻﻠﻴﺔ ﺑﻌﺪ إزاﺣﺔ اﻟﻤﺮآﺒﺔ اﻟﻤﻮﺳﻤﻴﺔ أي‬ yt = zt − St , t = 1, 2,..., n =Tt + Et , t = 1, 2,..., n

‫ أو اﻟﺒﻮاﻗﻲ ﺑﻌﺪ إزاﺣﺔ ﻣﺮآﺒﺘﻲ اﻹﻧﺠﺮاف‬Et ‫اﻟﺸﻜﻞ اﻷﺳﻔﻞ اﻷﻳﻤﻦ ﻳﻌﻄﻲ ﻣﺮآﺒﺔ اﻟﺨﻄﺄ‬ ‫واﻟﻤﻮﺳﻤﻴﺔ ﻣﻦ اﻟﻤﺸﺎهﺪات اﻷﺻﻠﻴﺔ أي‬ et = zt − Tt − St , t = 1, 2,..., n =Et , t = 1, 2,..., n

.‫ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﻣﻊ ﻣﻘﺎﻳﻴﺲ دﻗﺔ اﻟﺘﻄﺒﻴﻖ‬٢٤ ‫( ﻳﻌﻄﻲ اﻟﺘﻨﺒﺆات ﻟﻠﻘﻴﻢ‬٣) ‫ﺷﻜﻞ‬ zt = Tt St + Et , t = 1, 2,..., n

:‫ ﺗﻄﺒﻴﻖ اﻟﻨﻤﻮذج اﻟﺘﻀﺎﻋﻔﻲ‬:‫ﺛﺎﻧﻴﺎ‬

MTB > %Decomp 'GasDemand' 12; SUBC> Forecasts 24; SUBC> Title "Forecast of Gasoline Demand"; SUBC> Start 1. Executing from file: D:\MTBWIN\MACROS\Decomp.MAC Macro is running ... please wait

Time Series Decomposition Data Length NMissing

GasDeman 192.000 0

Trend Line Equation Yt = 96.4074 + 0.680579*t Seasonal Indices Period

Index

1 2 3 4 5 6 7 8 9 10 11 12

0.860355 0.828555 0.892431 0.936273 1.06124 1.07274 1.15775 1.17075 1.03409 1.05059 0.966300 0.968923

Accuracy of Model MAPE: MAD: MSD:

3.6338 5.7720 56.8996

Forecasts Row

Period

Forecast

1 2

193 194

195.954 189.275

١٥٧

3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216

204.474 215.156 244.596 247.977 268.415 272.227 241.154 245.718 226.660 227.935 202.980 196.042 211.762 222.803 253.263 256.738 277.870 281.789 249.599 254.298 234.552 235.848

(٤) ‫ﺷﻜﻞ‬ Forecast of Gasoline Demand Seasonal Indices

Original Data, by Seasonal Period

1.2

250

1.1

200

1.0 150 0.9 100 0.8 1 2 3 4 5 6 7 8 9 10 11 12

1 2 3 4 5 6 7 8 9 10 11 12

Percent Variation, by Seasonal Period 14 12 10 8 6 4 2 0

Residuals, by Seasonal Period 20 10 0 -10 -20 -30

1 2 3 4 5 6 7 8 9 10 11 12

1 2 3 4 5 6 7 8 9 10 11 12

(٥) ‫ﺷﻜﻞ‬ Forecast of Gasoline Demand Original Data

Detrended Data

250

1.3 1.2

200

1.1 1.0

150

0.9 100

0.8 0

100

200

0

Seasonally Adjusted Data 240 220 200 180 160 140

١٥٨

100

200

Seasonally Adj. and Detrended Data 20 10 0 -10

‫‪Forecast of Gasoline Demand‬‬

‫ﺷﻜﻞ )‪(٦‬‬ ‫‪Actual‬‬

‫‪280‬‬

‫‪Predicted‬‬ ‫‪Forecast‬‬

‫‪180‬‬

‫‪3.6338‬‬ ‫‪5.7720‬‬ ‫‪56.8996‬‬

‫‪MAPE:‬‬ ‫‪MAD:‬‬ ‫‪MSD:‬‬

‫‪GasDeman‬‬

‫‪Actual‬‬ ‫‪Predicted‬‬ ‫‪Forecast‬‬

‫‪80‬‬

‫‪200‬‬

‫‪100‬‬

‫‪0‬‬

‫‪Time‬‬

‫ﻣﻨﺎﻗﺸﺔ اﻟﻨﺘﺎﺋﺞ‪:‬‬ ‫اﻷﺷﻜﺎل )‪ (٤‬و )‪ (٥‬و )‪ (٦‬ﻟﻬﺎ ﻧﻔﺲ اﻟﺘﻔﺴﻴﺮ آﻤﺎ ﻓﻲ اﻷﺷﻜﺎل )‪ (١‬و )‪ (٢‬و )‪.(٣‬‬ ‫ﺑﻤﺎ اﻧﻨﺎ ﻃﺒﻘﻨﺎ ﻧﻤﻮذﺟﻴﻦ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﺸﺎهﺪة ﻓﻴﺠﺐ أن ﻧﺨﺘﺎر اﻓﻀﻞ ﻧﻤﻮذج‪ ،‬وهﻨﺎ ﻳﺄﺗﻲ دور‬ ‫ﻣﻘﺎﻳﻴﺲ دﻗﺔ اﻟﺘﻄﺒﻴﻖ واﻟﺘﻲ ﺗﻨﺘﺞ ﻣﻦ اﻟﺒﺮﻧﺎﻣﺞ‪ ،‬ﻟﺪﻳﻨﺎ ﺛﻼﺛﺔ ﻣﻘﺎﻳﻴﺲ دﻗﺔ‪:‬‬ ‫‪ -١‬ﻣﺘﻮﺳﻂ اﻟﺨﻄﺄ اﻟﻨﺴﺒﻲ اﻟﻤﻄﻠﻖ ‪ Mean Absolute Percentage Error‬أو ‪MAPE‬‬ ‫وﻳﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬ ‫‪zt − zˆt‬‬ ‫‪zt‬‬ ‫‪× 100, zt ≠ 0‬‬ ‫‪n‬‬

‫‪n‬‬

‫∑‬ ‫‪t =1‬‬

‫= ‪MAPE‬‬

‫‪ -٢‬ﻣﺘﻮﺳﻂ اﻹﻧﺤﺮاف اﻟﻤﻄﻠﻖ ‪ Mean Absolute Deviation‬أو ‪ MAD‬وﻳﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬ ‫‪n‬‬

‫‪− zˆt‬‬

‫‪t‬‬

‫‪∑z‬‬ ‫‪t =1‬‬

‫‪n‬‬

‫= ‪MAD‬‬

‫‪ -٣‬ﻣﺘﻮﺳﻂ أﻹﻧﺤﺮاف اﻟﻤﺮﺑﻊ )أو ﻣﺘﻮﺳﻂ اﻟﺨﻄﺄ اﻟﻤﺮﺑﻊ( ‪) MSD‬أو ‪ ( MSE‬وﻳﻌﻄﻰ ﺑﺎﻟﻌﻼﻗﺔ‬ ‫‪2‬‬

‫) ‪− zˆt‬‬

‫‪n‬‬

‫‪t‬‬

‫‪∑( z‬‬

‫‪n‬‬

‫‪t =1‬‬

‫= ‪MSD‬‬

‫ﺑﺈﺧﺘﻴﺎر أﺣﺪ هﺬة اﻟﻤﻘﺎﻳﻴﺲ ﻧﺨﺘﺎر اﻟﻨﻤﻮذج اﻟﺬي ﻳﻌﻄﻲ أﻗﻞ ﻗﻴﻤﺔ ﻟﻬﺬا اﻟﻤﻘﻴﺎس‪ ،‬اﻟﻤﻘﻴﺎس اﻷآﺜﺮ‬ ‫إﺳﺘﺨﺪاﻣﺎ وﺷﻴﻮﻋﺎ هﻮ ‪ MSD‬أو ‪ MSE‬وهﻮ اﻟﺬي ﺳﻮف هﻨﺎ‪.‬‬ ‫ﻟﻠﻨﻤﻮذج اﻹﺿﺎﻓﻲ ﻣﻘﺎﻳﻴﺲ اﻟﺪﻗﺔ هﻲ‪:‬‬ ‫‪3.6952‬‬ ‫‪١٥٩‬‬

‫‪MAPE:‬‬

‫‪MAD:‬‬ ‫‪MSD:‬‬

‫‪5.6622‬‬ ‫‪52.7851‬‬ ‫و ﻟﻠﻨﻤﻮذج اﻟﺘﻀﺎﻋﻔﻲ‪:‬‬

‫‪MAPE:‬‬ ‫‪3.6338‬‬ ‫‪MAD:‬‬ ‫‪5.7720‬‬ ‫‪MSD:‬‬ ‫‪56.8996‬‬ ‫ﻧﻼﺣﻆ أن اﻟﻨﻤﻮذج اﻹﺿﺎﻓﻲ اﻋﻄﻰ اﻗﻞ ﻗﻴﻤﺔ ﻟﻠﻤﻘﻴﺎس ‪ MSD‬وﻟﺬﻟﻚ ﻧﻘﺮر إﺳﺘﺨﺪام هﺬا اﻟﻨﻤﻮذج‬ ‫ﻟﻠﺘﻨﺒﺆ ﻋﻦ اﻟﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻄﻠﺐ‪.‬‬

‫ﺗﻮﺿﻴﺢ ﻃﺮﻳﻘﺔ ﺗﺤﻠﻴﻞ او ﺗﻔﻜﻴﻚ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻲ ﻣﺮآﺒﺎت ‪Decomposition‬‬ ‫‪:Method‬‬ ‫ﺳﻮف ﻧﻮﺿﺢ ﺑﺎﻟﻤﺜﺎل اﻟﺘﺎﻟﻲ ﻃﺮﻳﻘﺔ ﺗﺤﻠﻴﻞ أو ﺗﻔﻜﻴﻚ اﻟﻤﺘﺴﻠﺴﻠﺔ إﻟﻰ ﻣﺮآﺒﺎت‪ .‬اﻟﺒﻴﺎﻧﺎت اﻟﺘﺎﻟﻴﺔ إﻧﺘﺎج‬ ‫‪ ١٦٨‬ﻳﻮﻣﺎ ﻟﻠﺤﻠﻴﺐ ﻓﻲ أﺣﺪ اﻟﻤﺰارع ﺑﺎﻟﻜﻴﻠﻮ ﺟﺮام‬ ‫‪MTB > Read "E:\Mtbwin\milk.dat" c1.‬‬ ‫‪Entering data from file: E:\Mtbwin\milk.dat‬‬ ‫‪168 rows read.‬‬ ‫'‪MTB > name c1='MilkProd‬‬ ‫‪MTB > print c1‬‬

‫‪Data Display‬‬ ‫‪MilkProd‬‬ ‫‪582‬‬ ‫‪598‬‬ ‫‪634‬‬ ‫‪635‬‬ ‫‪688‬‬ ‫‪698‬‬ ‫‪711‬‬ ‫‪734‬‬ ‫‪751‬‬ ‫‪800‬‬ ‫‪805‬‬ ‫‪802‬‬ ‫‪813‬‬ ‫‪843‬‬

‫‪553‬‬ ‫‪565‬‬ ‫‪594‬‬ ‫‪602‬‬ ‫‪645‬‬ ‫‪660‬‬ ‫‪677‬‬ ‫‪690‬‬ ‫‪711‬‬ ‫‪763‬‬ ‫‪766‬‬ ‫‪760‬‬ ‫‪773‬‬ ‫‪797‬‬

‫‪577‬‬ ‫‪587‬‬ ‫‪611‬‬ ‫‪621‬‬ ‫‪667‬‬ ‫‪687‬‬ ‫‪706‬‬ ‫‪723‬‬ ‫‪747‬‬ ‫‪800‬‬ ‫‪810‬‬ ‫‪791‬‬ ‫‪812‬‬ ‫‪827‬‬

‫‪568‬‬ ‫‪583‬‬ ‫‪604‬‬ ‫‪615‬‬ ‫‪661‬‬ ‫‪681‬‬ ‫‪701‬‬ ‫‪725‬‬ ‫‪740‬‬ ‫‪790‬‬ ‫‪809‬‬ ‫‪784‬‬ ‫‪815‬‬ ‫‪817‬‬

‫‪599‬‬ ‫‪617‬‬ ‫‪639‬‬ ‫‪653‬‬ ‫‪697‬‬ ‫‪722‬‬ ‫‪740‬‬ ‫‪764‬‬ ‫‪783‬‬ ‫‪834‬‬ ‫‪855‬‬ ‫‪837‬‬ ‫‪867‬‬ ‫‪858‬‬

‫‪640‬‬ ‫‪660‬‬ ‫‪678‬‬ ‫‪702‬‬ ‫‪735‬‬ ‫‪767‬‬ ‫‪783‬‬ ‫‪801‬‬ ‫‪819‬‬ ‫‪869‬‬ ‫‪894‬‬ ‫‪881‬‬ ‫‪908‬‬ ‫‪896‬‬

‫‪697‬‬ ‫‪716‬‬ ‫‪736‬‬ ‫‪756‬‬ ‫‪798‬‬ ‫‪817‬‬ ‫‪826‬‬ ‫‪845‬‬ ‫‪859‬‬ ‫‪913‬‬ ‫‪935‬‬ ‫‪924‬‬ ‫‪947‬‬ ‫‪937‬‬

‫‪727‬‬ ‫‪742‬‬ ‫‪770‬‬ ‫‪782‬‬ ‫‪811‬‬ ‫‪837‬‬ ‫‪858‬‬ ‫‪871‬‬ ‫‪886‬‬ ‫‪942‬‬ ‫‪961‬‬ ‫‪957‬‬ ‫‪969‬‬ ‫‪966‬‬

‫‪640‬‬ ‫‪653‬‬ ‫‪688‬‬ ‫‪709‬‬ ‫‪736‬‬ ‫‪762‬‬ ‫‪775‬‬ ‫‪785‬‬ ‫‪807‬‬ ‫‪860‬‬ ‫‪890‬‬ ‫‪883‬‬ ‫‪889‬‬ ‫‪892‬‬

‫‪656‬‬ ‫‪673‬‬ ‫‪705‬‬ ‫‪722‬‬ ‫‪755‬‬ ‫‪784‬‬ ‫‪796‬‬ ‫‪805‬‬ ‫‪824‬‬ ‫‪878‬‬ ‫‪900‬‬ ‫‪898‬‬ ‫‪902‬‬ ‫‪903‬‬

‫‪561‬‬ ‫‪566‬‬ ‫‪618‬‬ ‫‪622‬‬ ‫‪635‬‬ ‫‪667‬‬ ‫‪696‬‬ ‫‪690‬‬ ‫‪707‬‬ ‫‪756‬‬ ‫‪799‬‬ ‫‪773‬‬ ‫‪778‬‬ ‫‪782‬‬

‫‪589‬‬ ‫‪600‬‬ ‫‪628‬‬ ‫‪658‬‬ ‫‪677‬‬ ‫‪713‬‬ ‫‪717‬‬ ‫‪734‬‬ ‫‪750‬‬ ‫‪804‬‬ ‫‪826‬‬ ‫‪821‬‬ ‫‪828‬‬ ‫‪834‬‬

‫وﻧﺮﺳﻢ اﻟﺒﻴﺎﻧﺎت اﻟﺴﺎﺑﻘﺔ ﺑﺎﻷواﻣﺮ‪:‬‬ ‫;'‪MTB > TSPlot 'MilkProd‬‬ ‫>‪SUBC‬‬ ‫;‪Index‬‬ ‫>‪SUBC‬‬ ‫;‪TDisplay 11‬‬ ‫>‪SUBC‬‬ ‫;‪Symbol‬‬ ‫>‪SUBC‬‬ ‫‪Connect.‬‬

‫‪1000‬‬

‫‪900‬‬

‫‪700‬‬

‫‪600‬‬

‫‪150‬‬

‫‪100‬‬

‫‪١٦٠‬‬

‫‪50‬‬

‫‪Index‬‬

‫‪MilkProd‬‬

‫‪800‬‬

:‫ ﺗﻄﺒﻴﻖ اﻟﻨﻤﻮذج اﻹﺿﺎﻓﻲ‬:‫اوﻻ‬ ‫ ﻟﻜﻲ‬zt = Tt + St + Et , t = 1, 2,..., n ‫ ﻧﻜﺘﺐ اﻟﻨﻤﻮذج ﻋﻠﻰ اﻟﺸﻜﻞ‬z1 , z2 ,..., zn ‫ﻟﻤﺸﺎهﺪات‬ :‫ﻧﻘﻮم ﺑﺘﻔﻜﻴﻚ هﺬﻩ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺰﻣﻨﻴﺔ اﻟﻲ اﻟﻤﺮآﺒﺎت اﻟﺴﺎﺑﻘﺔ ﺳﻮف ﻧﺴﺘﻌﺮض ﻃﺮﻳﻘﺘﻴﻦ ﻣﻤﻜﻨﺔ‬ :‫اﻟﻄﺮﻳﻘﺔ اﻻوﻟﻲ‬ :‫ أي‬Tt ‫ ﻧﻄﺒﻖ إﻧﺤﺪار ﺧﻄﻲ ﺑﺴﻴﻂ ﻟﻠﻤﺸﺎهﺪات ﻋﻠﻲ اﻟﺰﻣﻦ ﻟﺘﻘﺪﻳﺮ ﻣﺮآﺒﺔ اﻻﻧﺠﺮاف‬-١ Tˆt ≡ zˆt = a + bt , t = 1, 2,...,168

:‫أي‬ MTB > set c2 DATA> 1:168 DATA> end MTB > name c1='MilkProd' c2='Time' c3='Trend' c5='Detrend' c6='Index' c8='Fitted' c9='Resid' MTB > regr c1 1 c2; SUBC> fits c3.

Regression Analysis The regression equation is MilkProd = 612 + 1.69 Time Predictor Constant Time S = 60.74

Coef StDev T P 611.682 9.414 64.97 0.000 1.69262 0.09663 17.52 0.000 R-Sq = 64.9% R-Sq(adj) = 64.7%

Analysis of Variance Source Regression Error Total

DF 1 166 167

SS 1132003 612439 1744443

MS 1132003 3689

F 306.83

P 0.000

‫وﺷﻜﻞ اﻹﻧﺠﺮاف هﻮ‬

900

Trend

800

700

600

Index

50

100

150

‫ ﻧﻄﺮح ﻣﺮآﺒﺔ اﻹﻧﺠﺮاف ﻣﻦ اﻟﻤﺸﺎهﺪات اﻻﺻﻠﻴﺔ ﻓﻨﺤﺼﻞ ﻋﻠﻲ ﻣﺎﻳﺴﻤﻲ ﺑﺎﻟﻤﺘﺴﻠﺴﻠﺔ ﻣﺰاﻟﺔ‬-٢ zt − zˆt = zt − Tˆt , t = 1, 2,...,168 ‫ أي‬Detrended Series ‫اﻹﻧﺠﺮاف‬ MTB > let c5=c1-c3

:‫وﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‬

١٦١

Detrend

100

0

-100 Index

50

100

150

zt − Tˆt = St + Et , t = 1, 2,...,168 ‫ﻧﻼﺣﻆ اﻧﻬﺎ اﺻﺒﺤﺖ ﻣﻮﺳﻤﻴﺔ ﻓﻘﻂ ﻷن‬ :‫ آﺎﻟﺘﺎﻟﻲ‬Seasonal Indices ‫ ﻟﺘﻘﺪﻳﺮ اﻟﻤﺮآﺒﺔ اﻟﻤﻮﺳﻤﻴﺔ ﻧﻮﺟﺪ اﻟﻤﺆﺷﺮات اﻟﻤﻮﺳﻤﻴﺔ‬-٣ ‫ اﻟﻤﺆﺷﺮ اﻟﻤﻮﺳﻤﻲ ﻟﻠﺸﻬﺮ‬I1 ‫ ﺣﻴﺚ‬I s , s = 1, 2,...,12 ‫ﻟﻨﺮﻣﺰ ﻟﻠﻤﺆﺷﺮات اﻟﻤﻮﺳﻤﻴﺔ ﺑﺎﻟﺮﻣﺰ‬ d t = zt − Tˆt , t = 1, 2,...,168 ‫ اﻟﻤﺆﺷﺮ اﻟﻤﻮﺳﻤﻲ ﻟﻠﺸﻬﺮ اﻟﺜﺎﻧﻲ وهﻜﺬا وﻟﻨﺮﻣﺰ ﺑـ‬I 2 ‫اﻷول و‬ :‫ﺗﻘﺪر هﺬﻩ اﻟﻤﺆﺷﺮات آﺎﻟﺘﺎﻟﻲ‬

1 ( d1 + d13 + d 25 + L + d157 ) 14 1 I 2 = ( d 2 + d14 + d 26 + L + d158 ) 14 M I1 =

I12 =

1 ( d12 + d 24 + d 36 + L + d168 ) 14 :‫وﻳﺘﻢ ذﻟﻚ ﺑﺈﺳﺘﺨﺪام اﻷواﻣﺮ اﻟﺘﺎﻟﻴﺔ‬

MTB > DATA> DATA> MTB > SUBC> SUBC> MTB > & CONT> MTB > MTB > MTB > DATA> DATA> MTB >

set c4 14(1:12) end stat c5; by c4; mean c6. Stack 'Index' 'Index' 'Index' 'Index' 'Index' 'Index' 'Index' 'Index' 'Index' 'Index' 'Index' 'Index' 'Index' 'Index' c7. let c8=c3+c7 let c9=c1-c8 set c10 1:12 end print c10 c6

Data Display

١٦٢

‫‪ -٤‬اﻟﺘﻨﺒﺆات ﺗﻮﻟﺪ آﺎﻟﺘﺎﻟﻲ‪:‬‬

‫‪Index‬‬

‫‪Season‬‬

‫‪Row‬‬

‫‪-18.328‬‬ ‫‪-57.806‬‬ ‫‪34.716‬‬ ‫‪49.595‬‬ ‫‪110.616‬‬ ‫‪82.281‬‬ ‫‪32.517‬‬ ‫‪-9.747‬‬ ‫‪-52.297‬‬ ‫‪-48.775‬‬ ‫‪-79.754‬‬ ‫‪-43.018‬‬

‫‪1‬‬ ‫‪2‬‬ ‫‪3‬‬ ‫‪4‬‬ ‫‪5‬‬ ‫‪6‬‬ ‫‪7‬‬ ‫‪8‬‬ ‫‪9‬‬ ‫‪10‬‬ ‫‪11‬‬ ‫‪12‬‬

‫‪1‬‬ ‫‪2‬‬ ‫‪3‬‬ ‫‪4‬‬ ‫‪5‬‬ ‫‪6‬‬ ‫‪7‬‬ ‫‪8‬‬ ‫‪9‬‬ ‫‪10‬‬ ‫‪11‬‬ ‫‪12‬‬

‫‪z168 ( l ) = 612 + 1.69 ( l + 168) + I ( l mod 12 ) , l = 1,2,...‬‬

‫ﻓﻤﺜﻼ اﻟﺘﻨﺒﺆ ﻋﻨﺪ اﻟﻴﻮم ‪ ١٦٩‬هﻮ‬

‫‪z168 (1) = 612 + 1.69 (169 ) + I1‬‬ ‫‪=897.61 + ( −18.328 ) = 879.282‬‬

‫اﻟﻄﺮﻳﻘﺔ اﻟﺜﺎﻧﻴﺔ‪:‬‬ ‫وهﻲ اﻟﺘﻲ ﻳﺴﺘﺨﺪﻣﻬﺎ ﺑﺮﻧﺎﻣﺞ ‪: Minitab‬‬ ‫‪ -١‬آﺎﻟﻄﺮﻳﻘﺔ اﻻوﻟﻰ ﻧﻄﺒﻖ إﻧﺤﺪار ﺧﻄﻲ ﺑﺴﻴﻂ ﻟﻠﻤﺸﺎهﺪات ﻋﻠﻲ اﻟﺰﻣﻦ ﻟﺘﻘﺪﻳﺮ ﻣﺮآﺒﺔ اﻻﻧﺠﺮاف‬ ‫‪ Tt‬ﻓﻨﺤﺼﻞ ﻋﻠﻰ ﻧﻔﺲ اﻟﻨﺘﻴﺠﺔ آﻤﺎ ﻓﻲ اﻟﻄﺮﻳﻘﺔ اﻻوﻟﻰ )‪(١‬‬ ‫‪ -٢‬اﻳﻀﺎ هﻨﺎ ﻧﻄﺮح ﻣﺮآﺒﺔ اﻹﻧﺠﺮاف ﻣﻦ اﻟﻤﺸﺎهﺪات اﻻﺻﻠﻴﺔ ﻓﻨﺤﺼﻞ ﻋﻠﻲ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻣﺰاﻟﺔ‬ ‫اﻹﻧﺠﺮاف ‪Detrended Series‬‬ ‫‪ -٣‬ﻧﻄﺒﻖ اﻵن ﻣﺘﻮﺳﻂ ﻣﺘﺤﺮك ﻣﻦ درﺟﺔ اﻟﻤﻮﺳﻢ وﻧﻮﺳﻄﻪ اذا اﺣﺘﺎج اﻻﻣﺮ‬ ‫‪ -٤‬ﻧﻄﺮح اﻟﻤﺘﻮﺳﻄﺎت اﻟﻤﺘﺤﺮآﺔ ﻣﻦ ﻧﻈﻴﺮاﺗﻬﺎ ﻓﻲ اﻟﻤﺘﺴﻠﺴﻠﺔ ﻣﺰاﻟﺔ اﻹﻧﺠﺮاف ﻓﻨﺤﺼﻞ ﻋﻠﻲ‬ ‫ﻣﺘﺴﻠﺴﻠﺔ ﺗﺤﻮي اﻟﻤﺮآﺒﺎت اﻟﻤﻮﺳﻤﻴﺔ‬ ‫‪ -٥‬ﺗﻘﺪر اﻟﻤﺮآﺒﺎت اﻟﻤﻮﺳﻤﻴﺔ آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫) ‪I1 = Median ( d1 , d13 , d 25 ,L, d157‬‬

‫) ‪I 2 = Median ( d 2 , d14 , d 26 ,L, d158‬‬

‫‪M‬‬

‫) ‪I12 = Median ( d12 , d 24 , d 36 ,L, d168‬‬ ‫‪ -٦‬ﺗﻮﻟﺪ اﻟﺘﻨﺒﺆات آﺎﻟﺴﺎﺑﻖ‬ ‫وﺳﻮف ﻧﺴﺘﻌﺮض هﺬا آﺎﻟﺘﺎﻟﻲ‪:‬‬

‫‪MTB > Read "E:\Mtbwin\milk.dat" c1.‬‬ ‫‪Entering data from file: E:\Mtbwin\milk.dat‬‬ ‫‪168 rows read.‬‬ ‫'‪MTB > name c1='MilkProd‬‬ ‫‪MTB > set c2‬‬ ‫‪DATA> 1:168‬‬ ‫‪DATA> end‬‬ ‫'‪MTB > name c2='Time‬‬ ‫;‪MTB > regr c1 1 c2‬‬ ‫‪SUBC> fits c3.‬‬

‫‪Regression Analysis‬‬

‫‪١٦٣‬‬

The regression equation is MilkProd = 612 + 1.69 Time Predictor Constant Time

Coef 611.682 1.69262

S = 60.74

StDev 9.414 0.09663

R-Sq = 64.9%

T 64.97 17.52

P 0.000 0.000

R-Sq(adj) = 64.7%

Analysis of Variance Source Regression Error Total

DF 1 166 167

SS 1132003 612439 1744443

Unusual Observations Obs Time MilkProd 113 113 942.00 125 125 961.00

MS 1132003 3689

Fit 802.95 823.26

F 306.83

StDev Fit 5.44 6.11

P 0.000

Residual 139.05 137.74

St Resid 2.30R 2.28R

R denotes an observation with a large standardized residual MTB MTB MTB MTB

> > > >

name c3='Trend' let c4=c1-c3 name c4='Detrend' Name c5 = 'AVER1'

:‫ وﻧﻮﺳﻄﻪ‬١٢ ‫ﻓﻲ اﻟﺨﻄﻮة اﻟﺘﺎﻟﻴﺔ ﻧﻄﺒﻖ ﻣﺘﻮﺳﻂ ﻣﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ‬ MTB > %MA 'Detrend' 12; SUBC> Center; SUBC> Averages 'AVER1'. Executing from file: E:\MTBWIN\MACROS\MA.MAC Macro is running ... please wait

Moving average Data Length NMissing

Detrend 168.000 0

Moving Average Length: 12 Accuracy Measures MAPE: 111.68 MAD: 52.36 MSD: 3564.77

‫ﻧﻄﺮح اﻟﻤﺘﻮﺳﻄﺎت اﻟﻤﺘﺤﺮآﺔ اﻟﻤﻮﺳﻄﺔ ﻣﻦ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﺰال إﻧﺠﺮاﻓﻬﺎ‬ MTB > MTB > MTB > DATA> DATA> MTB > SUBC> SUBC> MTB >

let c6=c4-c5 name c6='DeSeason' set c2 14(1:12) end stat c6; by c2; median c7. name c7='SeasInx'

Data Display Row

Season

SeasInx

١٦٤

1 2 3 4 5 6 7 8 9 10 11 12

1 2 3 4 5 6 7 8 9 10 11 12

-20.750 -58.958 35.625 50.083 109.542 81.292 33.917 -10.000 -52.792 -50.250 -79.958 -44.375

:‫ﻧﻘﺎرن اﻟﺤﺴﺎﺑﺎت اﻟﺘﻲ اﺟﺮﻳﻨﺎهﺎ ﻣﻊ اﻟﺒﺮﻧﺎﻣﺞ اﻷﺻﻠﻲ‬ MTB > %Decomp 'MilkProd' 12; SUBC> Additive ; SUBC> Start 1. Executing from file: E:\MTBWIN\MACROS\Decomp.MAC Macro is running ... please wait

Time Series Decomposition Data Length NMissing

MilkProd 168.000 0

Trend Line Equation Yt = 611.682 + 1.69262*t Seasonal Indices Period 1 2 3 4 5 6 7 8 9 10 11 12

Index -20.1979 -58.4062 36.1771 50.6354 110.094 81.8437 34.4687 -9.44792 -52.2396 -49.6979 -79.4063 -43.8229

Accuracy of Model MAPE: MAD: MSD:

1.583 12.088 244.406

.‫وﺑﻤﻘﺎرﻧﺔ اﻟﻨﺘﻴﺠﺘﻴﻦ ﻧﺠﺪ اﻧﻬﻤﺎ ﺗﻘﺮﻳﺒﺎ ﻣﺘﺴﺎوﻳﺘﺎن‬ :‫ آﺎﻟﺘﺎﻟﻲ‬%Decomp ‫ﺳﻮف ﻧﻮﻟﺪ ﺗﻨﺒﺆات ﺑﺎﺳﺘﺨﺪام اﻟﺒﺮﻧﺎﻣﺞ‬ MTB > %Decomp 'MilkProd' 12; SUBC> Additive ; SUBC> Forecasts 12; SUBC> Start 1.

:‫واﻟﺘﻲ ﺗﻌﻄﻲ اﻟﺘﻨﺒﺆات‬ Forecasts

١٦٥

‫‪Forecast‬‬

‫‪Period‬‬

‫‪Row‬‬

‫‪877.54‬‬ ‫‪841.02‬‬ ‫‪937.30‬‬ ‫‪953.45‬‬ ‫‪1014.60‬‬ ‫‪988.04‬‬ ‫‪942.36‬‬ ‫‪900.13‬‬ ‫‪859.04‬‬ ‫‪863.27‬‬ ‫‪835.25‬‬ ‫‪872.53‬‬

‫‪169‬‬ ‫‪170‬‬ ‫‪171‬‬ ‫‪172‬‬ ‫‪173‬‬ ‫‪174‬‬ ‫‪175‬‬ ‫‪176‬‬ ‫‪177‬‬ ‫‪178‬‬ ‫‪179‬‬ ‫‪180‬‬

‫‪1‬‬ ‫‪2‬‬ ‫‪3‬‬ ‫‪4‬‬ ‫‪5‬‬ ‫‪6‬‬ ‫‪7‬‬ ‫‪8‬‬ ‫‪9‬‬ ‫‪10‬‬ ‫‪11‬‬ ‫‪12‬‬

‫ﺗﻤﺮﻳﻦ‪ :‬وﻟﺪ ﺗﻨﺒﺆات ﻹﻧﺘﺎج اﻟﺤﻠﻴﺐ اﻟﻴﻮﻣﻲ ﺑﺈﺳﺘﺨﺪام اﻟﻄﺮﻳﻘﺘﻴﻦ اﻟﻤﻌﻄﺎة وﻗﺎرﻧﻬﺎ ﺑﺎﻟﻘﻴﻢ اﻷﺧﻴﺮة‬ ‫اﻟﻨﺎﺗﺠﺔ ﻣﻦ اﻟﺒﺮﻧﺎﻣﺞ‬

‫‪١٦٦‬‬

‫اﻟﻔﺼﻞ اﻟﻌﺎﺷﺮ‬ ‫اﻟﺘﻤﻬﻴﺪ و اﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ‪Using Moving‬‬ ‫‪Average Smoothing for Forecasting‬‬ ‫ﻳﺴﺘﺨﺪم اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻟﺘﻤﻬﻴﺪ اﻟﻤﺸﺎهﺪات وذﻟﻚ ﺑﺘﻘﻠﻴﻞ ﺗﺒﺎﻳﻦ اﻷﺧﻄﺎء ﻓﻤﺜﻼ ﻟﻮ آﺎن ﻟﺪﻳﻨﺎ‬ ‫ﻣﺸﺎهﺪات ﻣﻦ ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ‪ z1 , z2 , z3 ,K , zn −2 , zn −1 , zn‬ﻓﺎﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ ‪m‬‬ ‫ﻟﻠﻤﺸﺎهﺪات ﻳﺤﺴﺐ ﻣﻦ اﻟﻌﻼﻗﺔ‬ ‫‪1‬‬ ‫‪zˆt = ( zt + zt −1 + zt −2 + L + zt −m+1 ) , t = m, m + 1,..., n‬‬ ‫‪m‬‬ ‫أو‬ ‫‪1‬‬ ‫‪zˆt = zˆt −1 + ( zt − zt −m ) , t = m, m + 1,..., n‬‬ ‫‪m‬‬ ‫ﻻﺣﻆ ان ﻋﺪد اﻟﻤﺸﺎهﺪات اﺻﺒﺢ ﺑﻌﺪ اﻟﺘﻤﻬﻴﺪ ‪. n − m + 1‬‬ ‫ﻓﻤﺜﻼ ﻟﻮ آﺎﻧﺖ ‪ m=٣‬ﻓﺈن اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ اﻟﺜﺎﻟﺜﺔ هﻮ‬ ‫‪1‬‬ ‫) ‪( z3 + z2 + z1‬‬ ‫‪3‬‬ ‫‪1‬‬ ‫‪1‬‬ ‫) ‪zˆ4 = ( z4 + z3 + z2 ) or zˆ4 = zˆ3 + ( z4 − z1‬‬ ‫‪3‬‬ ‫‪3‬‬ ‫‪M‬‬ ‫‪1‬‬ ‫‪1‬‬ ‫) ‪zˆn = ( zn + zn −1 + zn −2 ) or zˆn = zˆn −1 + ( zn − zn −3‬‬ ‫‪3‬‬ ‫‪3‬‬ ‫= ‪zˆ3‬‬

‫وﻟﻜﻲ ﻧﺮى آﻴﻒ ﻳﻌﻤﻞ اﻟﺘﻤﻬﻴﺪ ﻟﺘﻘﻠﻴﻞ ﺗﺒﺎﻳﻦ اﻷﺧﻄﺎء ﻟﻨﻔﺘﺮض ان اﻟﻤﺸﺎهﺪات ﺗﺘﺒﻊ اﻟﻨﻤﻮذج‬ ‫‪WN ( 0, σ 2 ) , t = 1, 2,..., n‬‬

‫ﻓﻴﻜﻮن‬

‫‪zt = µ + at , at‬‬

‫‪V ( zt ) = σ , ∀t‬‬ ‫‪2‬‬

‫وﺑﺎﻟﺘﺎﻟﻲ‬ ‫‪2‬‬

‫‪σ‬‬

‫= ) ‪V ( zˆt‬‬

‫‪, t = m, m + 1,..., n‬‬ ‫‪m‬‬ ‫أي ان اﻟﻤﺸﺎهﺪات اﻟﻤﻤﻬﺪة اﺻﺒﺢ ﺗﺒﺎﻳﻨﻬﺎ أﺻﻐﺮ ﺑـ ‪ m‬ﺿﻌﻒ ﻣﻦ اﻟﻤﺸﺎهﺪات اﻷﺻﻠﻴﺔ وهﺬا اﻟﺘﻤﻬﻴﺪ‬ ‫ﻟﻸﺧﻄﺎء ﻳﻈﻬﺮ أي ﻧﻤﻂ ﻓﻲ اﻟﻤﺘﺴﻠﺴﻠﺔ آﺎن ﻣﺪﻓﻮﻧﺎ او ﻣﻐﻄﻰ ﻣﻦ ﺗﺄﺛﻴﺮ اﻷﺧﻄﺎء‪.‬‬ ‫ﻣﻼﺣﻈﺔ‪ :‬ﺗﺆﺧﺬ ‪ m‬داﺋﻤﺎ ﻓﺮدﻳﺔ وذﻟﻚ ﻟﻨﺘﺠﻨﺐ ﺗﻮﺳﻴﻂ اﻟﻘﻴﻢ اﻟﻤﻤﻬﺪة‪.‬‬

‫اﻟﺘﻨﺒﺆ ﺑﺈﺳﺘﺨﺪام اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك‪:‬‬ ‫ﻳﺆﺧﺬ آﻤﺘﻨﺒﺊ ﻟﻠﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك‪:‬‬

‫‪zn ( l ) = zˆn −1 , l > 0‬‬

‫ﻣﺜﺎل‪:‬‬ ‫ﺑﺈﺳﺘﺨﺪام اﻟﺤﺰﻣﺔ اﻷﺣﺼﺎﺋﻴﺔ ‪ MINITAB‬ﻧﺤﻤﻞ اﻟﺒﻴﺎﻧﺎت ﻣﻦ ورﻗﺔ اﻟﻌﻤﻞ ‪EMPLOY.MTW‬‬ ‫‪'E:\Mtbwin\DATA\EMPLOY.MTW'.‬‬

‫‪MTB > Retrieve‬‬

‫ﻧﻨﻈﺮ ﻣﺎذا ﺗﺤﻮي ﻣﻦ ﻣﺘﻐﻴﺮات‬ ‫‪MTB > info‬‬

‫‪١٦٧‬‬

Information on the Worksheet Column C1 C2 C3

Count 60 60 60

Name Trade Food Metals

Metals ‫ﺳﻮف ﻧﺴﺘﺨﺪم اﻟﻤﺸﺎهﺪات ﻟﻠﻤﺘﻐﻴﺮ‬ Metals 44.2 43.1 42.4 42.9 45.0 49.4 50.5

44.3 42.6 43.1 43.6 45.5 50.0 51.2

44.4 42.4 43.2 44.7 46.2 50.0 50.7

43.4 42.2 42.8 44.5 46.8 49.6 50.3

42.8 41.8 43.0 45.0 47.5 49.9 49.2

44.3 40.1 42.8 44.8 48.3 49.6 48.1

44.4 42.0 42.5 44.9 48.3 50.7

44.8 42.4 42.6 45.2 49.1 50.7

44.4 43.1 42.3 45.2 48.9 50.9

:‫ﻧﺮﺳﻢ هﺬﻩ اﻟﻤﺸﺎهﺪات‬ MTB > TSPlot 'Metals'; SUBC> Index; SUBC> TDisplay 11; SUBC> Symbol; SUBC> Connect.

Metals

50

45

40 Index

10

20

30

40

50

60

‫ وﻧﻮﺟﺪ ﺗﻨﺒﺆات‬m=٣ ‫ﻧﻄﺒﻖ اﻵن ﺗﻤﻬﻴﺪا ﻟﻬﺬﻩ اﻟﻤﺸﺎهﺪات ﺑﺈﺳﺘﺨﺪام اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ‬ :‫ ﻗﻴﻢ ﻣﺴﺘﻘﺒﻠﻴﻪ‬٦ ‫ﻟـ‬ MTB > %MA 'Metals' 3; SUBC> Forecasts 6; SUBC> Title "Smoothing and Forecasting Metals". Executing from file: E:\MTBWIN\MACROS\MA.MAC

Moving average Data Length NMissing

Metals 60.0000 0

Moving Average Length: 3 Accuracy Measures MAPE: 1.55036

١٦٨

‫‪0.70292‬‬ ‫‪0.76433‬‬ ‫‪Lower‬‬

‫‪Upper‬‬

‫‪50.9135‬‬ ‫‪50.9135‬‬ ‫‪50.9135‬‬ ‫‪50.9135‬‬ ‫‪50.9135‬‬ ‫‪50.9135‬‬

‫‪Forecast‬‬

‫‪Period‬‬

‫‪49.2‬‬ ‫‪49.2‬‬ ‫‪49.2‬‬ ‫‪49.2‬‬ ‫‪49.2‬‬ ‫‪49.2‬‬

‫‪61‬‬ ‫‪62‬‬ ‫‪63‬‬ ‫‪64‬‬ ‫‪65‬‬ ‫‪66‬‬

‫‪47.4865‬‬ ‫‪47.4865‬‬ ‫‪47.4865‬‬ ‫‪47.4865‬‬ ‫‪47.4865‬‬ ‫‪47.4865‬‬

‫‪MAD:‬‬ ‫‪MSD:‬‬

‫‪Row‬‬ ‫‪1‬‬ ‫‪2‬‬ ‫‪3‬‬ ‫‪4‬‬ ‫‪5‬‬ ‫‪6‬‬

‫‪Smoothing and Forecasting Metals‬‬

‫‪Actual‬‬ ‫‪Predicted‬‬

‫‪50‬‬

‫‪Forecast‬‬

‫‪45‬‬ ‫‪Moving Average‬‬ ‫‪3‬‬

‫‪Metals‬‬

‫‪Actual‬‬ ‫‪Predicted‬‬ ‫‪Forecast‬‬

‫‪Length:‬‬

‫‪MAPE: 1.55036‬‬ ‫‪0.70292‬‬

‫‪MAD:‬‬

‫‪0.76433‬‬

‫‪MSD:‬‬

‫‪40‬‬ ‫‪60‬‬

‫ﺛﺎﻧﻴﺎ‪ :‬ﻣﻨﺎﻗﺸﺔ اﻟﻨﺘﺎﺋﺞ‬ ‫ﻓﻲ اﻟﻤﺜﺎل اﻟﺤﺎﻟﻲ‬

‫‪50‬‬

‫‪40‬‬

‫‪30‬‬

‫‪20‬‬

‫‪10‬‬

‫‪0‬‬

‫‪Time‬‬

‫‪50.3 + 49.2 + 48.1 147.6‬‬ ‫=‬ ‫‪= 49.2‬‬ ‫‪3‬‬ ‫‪3‬‬ ‫ﺗﺆﺧﺬ اﻟﺘﻨﺒﺆات ﻟﻠﻘﻴﻢ اﻟـ ‪ ٦‬اﻟﻤﺴﺘﻘﺒﻠﻴﺔ أي ﻟﻠﻘﻴﻢ ‪ zn +1 , zn +2 ,..., zn+6‬أو ﻓﻲ هﺬا اﻟﻤﺜﺎل‬ ‫‪ z61 , z62 ,..., z66‬آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫‪z60 (1) = z60 ( 2 ) = L = z60 ( 6 ) = 49.2‬‬

‫= ‪zˆ59‬‬

‫ﻟﺤﺴﺎب ﻓﺘﺮات ﺗﻨﺒﺆ ‪ 95%‬ﻧﺤﺴﺐ اﻟﻜﻤﻴﺎت ‪ ⎡⎣ zn ( l ) ± 1.96σˆ ⎤⎦ , l > 0‬أي‬ ‫] ˆ‪ [ 49.2 ± 1.96σ‬ﻟﺠﻤﻴﻊ ﻗﻴﻢ اﻟﺘﻨﺒﺆات اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ ،‬ﻧﺄﺧﺬ اﻟﻘﻴﻤﺔ ‪ MSD = 0.76433‬آﻤﻘﺪر ﻟـ‬

‫‪ σ 2‬أي ‪ σˆ 2 = 0.76433‬ﻓﻴﻜﻮن ‪ σˆ = 0.8743‬وﻋﻠﻴﻪ ﺗﻜﻮن ﻓﺘﺮة ﺗﻨﺒﺆ ‪ 95%‬ﻟﺠﻤﻴﻊ اﻟﻘﻴﻢ‬ ‫اﻟﻤﺴﺘﻘﺒﻠﻴﺔ هﻲ‪:‬‬ ‫]‪⎡⎣ 49.2 ± 1.96 ( 0.8743) ⎤⎦ = [ 49.2 ± 1.7135] = [ 47.4865,50.9135‬‬ ‫أي‪:‬‬ ‫‪z60+l ∈ [ 47.4865,50.9135] , l > 0 with probability 0.95‬‬ ‫ﻣﻼﺣﻈﺔ‪ :‬ﺗﺤﺴﺐ ‪ MSD‬آﺎﻵﺗﻲ‬

‫‪١٦٩‬‬

‫) ‪− zˆi‬‬

‫‪n −1‬‬

‫‪∑( z‬‬

‫‪i‬‬

‫‪i =2‬‬

‫‪n−2‬‬

‫= ˆ‪MSD = σ‬‬ ‫‪2‬‬

‫ﺗﻤﺮﻳﻦ‪:‬‬ ‫ﻃﺒﻖ ﻣﺘﻮﺳﻄﺎت ﻣﺘﺤﺮآﺔ ﻣﻦ اﻟﺪرﺟﺎت ‪ ٥‬و ‪ ٧‬ﻋﻠﻲ اﻟﻤﺸﺎهﺪات اﻟﺴﺎﺑﻘﺔ وﻗﺮر اﻳﻬﺎ اﻓﻀﻞ ﻟﻠﺘﻨﺒﺆ‬ ‫ﻋﻦ اﻟﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ؟‪.‬‬ ‫اﻟﻮﺳﻴﻂ اﻟﺠﺎري ‪Running Median‬‬ ‫اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻳﺘﺄﺛﺮ آﺜﻴﺮا ﺑﺎﻟﻤﺸﺎهﺪات اﻟﺨﺎرﺟﺔ ‪ Outliers‬او اﻟﻤﺘﻄﺮﻓﺔ ﻓﺎﻟﻘﻴﻤﺔ اﻟﻤﺘﻄﺮﻓﺔ‬ ‫اﻟﻮاﺣﺪة ﺗﺆﺛﺮ ﻋﻠﻲ ‪ m‬ﻣﻦ اﻟﻤﺘﻮﺳﻄﺎت اﻟﻤﺘﺤﺮآﺔ اﻟﻤﺘﺘﺎﻟﻴﺔ ﻓﻤﺜﻼ ﻟﻮآﺎﻧﺖ ﻟﺪﻳﻦ اﻟﻤﺸﺎهﺪات‬ ‫‪18‬‬

‫‪13‬‬

‫‪15‬‬

‫‪11‬‬

‫‪12‬‬

‫‪1500‬‬

‫‪10‬‬

‫‪6‬‬

‫‪8‬‬

‫‪9‬‬

‫)‪z(t‬‬ ‫‪5‬‬ ‫‪7‬‬ ‫‪20‬‬

‫‪3‬‬

‫وﻟﻬﺎ اﻟﺸﻜﻞ‬ ‫‪1500‬‬

‫‪1000‬‬

‫)‪z(t‬‬ ‫‪500‬‬

‫‪0‬‬ ‫‪10‬‬

‫‪In d e x‬‬

‫‪5‬‬

‫ﺑﺄﺧﺬ ﻣﺘﻮﺳﻂ ﻣﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ ‪ ٣‬ﻧﺠﺪ‬ ‫‪M o v in g A v e r a g e‬‬

‫‪A c tu a l‬‬

‫‪1500‬‬

‫‪P re d ic te d‬‬ ‫‪A c tu a l‬‬ ‫‪P re d ic te d‬‬

‫‪1000‬‬

‫)‪z(t‬‬

‫‪M o v in g A v e ra g e‬‬ ‫‪3‬‬ ‫‪1081‬‬

‫‪500‬‬

‫‪L e n g th :‬‬ ‫‪M APE :‬‬

‫‪273‬‬

‫‪M AD :‬‬

‫‪268811‬‬

‫‪MSD:‬‬

‫‪0‬‬

‫‪15‬‬

‫‪5‬‬

‫‪10‬‬

‫‪0‬‬

‫‪T im e‬‬

‫ﻻﺣﻆ ان اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك اﻟﻨﺎﺗﺞ ﺗﺄﺛﺮت ﻓﻴﺔ ﺛﻼﺛﺔ ﻗﻴﻢ ﺑﺎﻟﻘﻴﻤﺔ اﻟﻤﺘﻄﺮﻓﺔ‪.‬‬ ‫ﻟﻠﺘﻐﻠﺐ ﻋﻠﻰ ﻣﺜﻞ هﺬﻩ اﻟﺼﻌﻮﺑﺎت ﻳﺴﺘﺨﺪم اﻟﻮﺳﻴﻂ اﻟﺠﺎري ذا اﻟﻄﻮل اﻟﻔﺮدي آﻤﻤﻬﺪ ﻏﻴﺮ ﺧﻄﻲ‬ ‫واﻟﺬي ﻻﻳﺘﺄﺛﺮ ﺑﺎﻗﻴﻢ اﻟﻤﺘﻄﺮﻓﺔ‪.‬‬ ‫اﻟﻮﺳﻴﻂ اﻟﺠﺎري ذا اﻟﻄﻮل اﻟﻔﺮدي ‪ j = 2i + 1‬ﻟﻤﺸﺎهﺪات ‪ z1 , z2 , z3 ,K , zn −2 , zn −1 , zn‬ﻳﺤﺴﺐ‬ ‫ﻣﻦ اﻟﻌﻼﻗﺔ‬

‫‪z%t = med ( zt −i ,..., zt ,..., zt +i ) , j = 2i + 1‬‬

‫‪١٧٠‬‬

‫ﻓﻤﺜﻼ ﻟﻘﻴﻤﺔ ‪ j = 3‬ﺗﺼﺒﺢ اﻟﻌﻼﻗﺔ ) ‪ z%t = med ( zt −1 , zt , zt +1‬وﺑﺄﺧﺬ وﺳﻴﻂ ﺟﺎري ذا اﻟﻄﻮل ‪٣‬‬ ‫ﻟﻠﻤﺸﺎهﺪات اﻟﺴﺎﺑﻘﺔ ﻧﺠﺪ‬

‫‪15‬‬

‫)‪smoothz(t‬‬

‫‪10‬‬

‫‪5‬‬

‫‪12‬‬

‫‪8‬‬

‫‪10‬‬

‫‪4‬‬

‫‪6‬‬

‫‪2‬‬

‫‪In d e x‬‬

‫واذا آﺎﻧﺖ اﻟﻘﻴﻤﺔ اﻟﺤﻘﻴﻘﻴﺔ ﻟـ ‪ z9‬هﻲ ‪ ١٥‬وﻟﻴﺲ ‪ ١٥٠٠‬ﻓﺈن اﻟﻤﺸﺎهﺪات اﻟﺤﻘﻴﻘﻴﺔ ﻟﻬﺎ اﻟﺸﻜﻞ اﻟﺘﺎﻟﻲ‬

‫‪20‬‬

‫‪15‬‬

‫)‪z(t‬‬ ‫‪10‬‬

‫‪5‬‬

‫‪10‬‬

‫ﻗﺎرن ﺑﻴﻦ اﻟﻨﺘﻴﺠﺘﻴﻦ‪.‬‬

‫‪١٧١‬‬

‫‪5‬‬

‫‪Index‬‬

‫اﻟﻔﺼﻞ اﻟﺤﺎدي ﻋﺸﺮ‬ ‫اﻟﺘﻤﻬﻴﺪ و اﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ اﻟﺘﻤﻬﻴﺪ اﻻﺳﻲ اﻟﺒﺴﻴﻂ ‪Using Single‬‬ ‫‪: Exponential Smoothing for Forecasting‬‬ ‫اﻟﺘﻤﻬﻴﺪ ﺑﻮاﺳﻄﺔ اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻳﻌﻄﻲ ﺟﻤﻴﻊ اﻟﺒﻴﺎﻧﺎت ﻧﻔﺲ اﻷهﻤﻴﺔ وﺑﺎﻟﺘﺎﻟﻲ ﻓﺈن اﻟﻘﻴﻢ اﻟﻘﺪﻳﻤﺔ‬ ‫ﻧﻮﻋﺎ ﺗﺆﺛﺮ ﻧﻔﺲ اﻟﺘﺄﺛﻴﺮ آﺎﻟﻘﻴﻢ اﻟﺤﺪﻳﺜﺔ وهﺬا ﻗﺪ ﻻﻳﻜﻮن ﻣﻦ اﻟﻨﺎﺣﻴﺔ اﻟﻌﻤﻠﻴﺔ ﺻﺤﻴﺤﺎ‪ ،‬اﻟﺘﻤﻬﻴﺪ اﻻﺳﻲ‬ ‫ﻋﻠﻰ اﻟﻌﻜﺲ ﻳﻌﻄﻲ اﻟﻘﻴﻢ اﻷآﺜﺮ‬ ‫ﺣﺪاﺛﺔ أهﻤﻴﺔ أآﺒﺮ واﻟﻘﻴﻢ اﻻﺧﺮي ﺗﻌﻄﻰ اهﻤﻴﺔ ﺗﺘﻨﺎﻗﺺ اﺳﻴﺎ ﻣﻊ ﻗﺪﻣﻬﺎ‪ .‬ﻓﻤﺜﻼ ﻟﻮ آﺎن ﻟﺪﻳﻨﺎ‬ ‫ﻣﺸﺎهﺪات ﻣﻦ ﻣﺘﺴﻠﺴﻠﺔ زﻣﻨﻴﺔ ‪ z1 , z2 , z3 ,K , zn −2 , zn −1 , zn‬ﻓﺎﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻣﻦ اﻟﺪرﺟﺔ ‪m‬‬ ‫ﻟﻠﻤﺸﺎهﺪات ﻳﺤﺴﺐ ﻣﻦ اﻟﻌﻼﻗﺔ‬ ‫‪1‬‬ ‫‪zˆt = ( zt + zt −1 + zt −2 + L + zt −m+1 ) , t = m, m + 1,..., n‬‬ ‫‪m‬‬ ‫واﻟﺘﻲ ﻳﻤﻜﻦ آﺘﺎﺑﺘﻬﺎ‬ ‫‪1‬‬ ‫‪1‬‬ ‫‪1‬‬ ‫‪1‬‬ ‫‪zˆt = zt + zt −1 + zt −2 + L + zt −m+1 , t = m, m + 1,..., n‬‬ ‫‪m‬‬ ‫‪m‬‬ ‫‪m‬‬ ‫‪m‬‬ ‫‪1‬‬ ‫= ‪zˆt = β zt + β zt −1 + β zt −2 + L + β zt −m+1 , t = m, m + 1,..., n, β‬‬ ‫‪m‬‬ ‫أي ان اﻟﻤﺘﻮﺳﻂ اﻟﻤﺘﺤﺮك ﻳﻌﻄﻲ ﺟﻤﻴﻊ اﻟﺒﻴﺎﻧﺎت ﻧﻔﺲ اﻟﻮزن ‪β‬‬ ‫اﻵن ﻟﻮ أﻋﻄﻴﻨﺎ اﻟﺒﻴﺎﻧﺎت اوزان ﺗﺘﻨﺎﻗﺺ اﺳﻴﺎ ﻣﻊ ﺑُﻌﺪ اﻟﻤﺸﺎهﺪات ﻋﻦ اﻟﻘﻴﻤﺔ اﻟﺤﺎﺿﺮة ‪ zn‬آﺎﻟﺘﺎﻟﻲ‬ ‫‪st = α zt + α (1 − α ) zt −1 + α (1 − α ) zt −2 + L , t = 1, 2,..., n, 0 < α < 1‬‬ ‫‪2‬‬

‫اﻟﻘﻴﻤﺔ ‪ st‬هﻲ ﻣﺘﻮﺳﻂ ﻣﻮزون ﺑﺄوزان ﺗﺘﻨﺎﻗﺺ اﺳﻴﺎ ﻟﺠﻤﻴﻊ اﻟﻘﻴﻢ اﻟﺴﺎﺑﻘﺔ وهﺬا ﻣﺎﻳﺴﻤﻰ ﺑﺎﻟﺘﻤﻬﻴﺪ‬ ‫اﻻﺳﻲ اﻟﺒﺴﻴﻂ وﻳﻜﺘﺐ ﺑﺸﻜﻞ ﺗﻜﺮاري‬

‫‪st = α zt + (1 − α ) st −1 , t = 1, 2,..., n, s0 = z‬‬

‫وﺗﺆﺧﺬ اﻟﺘﻨﺒﺆات‬

‫‪zn ( l ) = sn , l ≥ 1‬‬

‫ﻣﺜﺎل‪:‬‬

‫ﺗﺤﻤﻞ اﻟﺒﻴﺎﻧﺎت ﻣﻦ ورﻗﺔ اﻟﻌﻤﻞ‬

‫‪EMPLOY.MTW‬‬ ‫‪'E:\Mtbwin\DATA\EMPLOY.MTW'.‬‬

‫‪MTB > Retrieve‬‬

‫ﺳﻮف ﻧﺴﺘﺨﺪم اﻟﻤﺸﺎهﺪات ﻓﻲ اﻟﺘﻐﻴﺮ ‪Metals‬‬ ‫‪44.4‬‬ ‫‪43.1‬‬ ‫‪42.3‬‬ ‫‪45.2‬‬ ‫‪48.9‬‬ ‫‪50.9‬‬

‫‪44.8‬‬ ‫‪42.4‬‬ ‫‪42.6‬‬ ‫‪45.2‬‬ ‫‪49.1‬‬ ‫‪50.7‬‬

‫‪44.4‬‬ ‫‪42.0‬‬ ‫‪42.5‬‬ ‫‪44.9‬‬ ‫‪48.3‬‬ ‫‪50.7‬‬

‫‪44.3‬‬ ‫‪40.1‬‬ ‫‪42.8‬‬ ‫‪44.8‬‬ ‫‪48.3‬‬ ‫‪49.6‬‬ ‫‪48.1‬‬

‫‪42.8‬‬ ‫‪41.8‬‬ ‫‪43.0‬‬ ‫‪45.0‬‬ ‫‪47.5‬‬ ‫‪49.9‬‬ ‫‪49.2‬‬

‫‪43.4‬‬ ‫‪42.2‬‬ ‫‪42.8‬‬ ‫‪44.5‬‬ ‫‪46.8‬‬ ‫‪49.6‬‬ ‫‪50.3‬‬

‫‪44.4‬‬ ‫‪42.4‬‬ ‫‪43.2‬‬ ‫‪44.7‬‬ ‫‪46.2‬‬ ‫‪50.0‬‬ ‫‪50.7‬‬

‫‪44.3‬‬ ‫‪42.6‬‬ ‫‪43.1‬‬ ‫‪43.6‬‬ ‫‪45.5‬‬ ‫‪50.0‬‬ ‫‪51.2‬‬

‫‪Metals‬‬ ‫‪44.2‬‬ ‫‪43.1‬‬ ‫‪42.4‬‬ ‫‪42.9‬‬ ‫‪45.0‬‬ ‫‪49.4‬‬ ‫‪50.5‬‬

‫ﻧﺮﺳﻢ هﺬﻩ اﻟﻤﺸﺎهﺪات‪:‬‬ ‫;'‪MTB > TSPlot 'Metals‬‬ ‫>‪SUBC‬‬ ‫;‪Index‬‬ ‫>‪SUBC‬‬ ‫;‪TDisplay 11‬‬ ‫>‪SUBC‬‬ ‫;‪Symbol‬‬

‫‪١٧٢‬‬

SUBC>

Connect.

M etals

50

45

40 Index

10

20

30

40

50

60

α = 0.2 ‫ﻧﻄﺒﻖ اﻵن ﺗﻤﻬﻴﺪا ﻟﻬﺬﻩ اﻟﻤﺸﺎهﺪات ﺑﺈﺳﺘﺨﺪام اﻟﺘﻤﻬﻴﺪ اﻻﺳﻲ اﻟﺒﺴﻴﻂ وﻧﺄﺧﺬ ﺛﺎﺑﺖ ﺗﻤﻬﻴﺪ‬ :‫ ﻗﻴﻢ ﻣﺴﺘﻘﺒﻠﻴﻪ‬٦ ‫وﻧﻮﺟﺪ ﺗﻨﺒﺆات ﻟـ‬ MTB > %SES 'Metals'; SUBC> Weight 0.2; SUBC> Forecasts 6; SUBC> Title "Smoothing and Forecasting Metals"; SUBC> Initial 6. Single Exponential Smoothing Data Metals Length 60.0000 NMissing 0 Smoothing Constant Alpha: 0.2 Accuracy Measures MAPE: 2.17304 MAD: 1.00189 MSD: 1.45392 Row

Period

Forecast

Lower

Upper

1 2 3 4 5 6

61 62 63 64 65 66

49.7216 49.7216 49.7216 49.7216 49.7216 49.7216

47.2670 47.2670 47.2670 47.2670 47.2670 47.2670

52.1763 52.1763 52.1763 52.1763 52.1763 52.1763

Smoothing and Forecasting Metals

Actual Predicted Forecast

Metals

50

Actual Predicted Forecast

45

Smoothing Constant Alpha:

40 0

10

20

30

Time

١٧٣

40

50

60

0.200

MAPE:

2.17304

MAD:

1.00189

MSD:

1.45392

‫ﺛﺎﻧﻴﺎ‪ :‬ﻣﻨﺎﻗﺸﺔ اﻟﻨﺘﺎﺋﺞ‬ ‫‪ -١‬ﻳﺤﺴﺐ اﻟﺘﻤﻬﻴﺪ اﻻﺳﻲ اﻟﺒﺴﻴﻂ ﻟﻤﺸﺎهﺪات‬ ‫ﻣﻦ اﻟﻌﻼﻗﺔ اﻟﺘﻜﺮارﻳﺔ‪:‬‬

‫‪ z1 , z2 ,K , zn −1 , zn −2‬ﺑﺜﺎﺑﺖ ﺗﻤﻬﻴﺪ ‪α = 0.2‬‬

‫‪si = α zi + (1 − α ) si −1 , i = 1, 2,..., n‬‬ ‫ﻟﻜﻲ ﻧﺒﺪأ اﻟﻌﻼﻗﺔ اﻟﺘﻜﺮارﻳﺔ ﻟﺤﺴﺎب اﻟﻘﻴﻢ اﻟﻤﻤﻬﺪة اﺳﻴﺎ ﻧﺤﺘﺎج اﻟﻲ اﻟﻘﻴﻤﺔ اﻻوﻟﻴﺔ ‪ s0‬واﻟﺘﻲ ﺗﺤﺴﺐ‬ ‫ﺑﻌﺪة ﻃﺮق‪ ،‬أﺣﺪ هﺬﻩ اﻟﻄﺮق واﻟﺘﻲ ﺳﻨﺴﺘﺨﺪﻣﻬﺎ هﻲ وﺿﻊ ‪ s0‬ﻣﺴﺎوﻳﺔ ﻟﻠﻤﺘﻮﺳﻂ‬ ‫‪m‬‬

‫) ‪, m = 6 ( or n, if n<6‬‬

‫‪∑z‬‬

‫‪i‬‬

‫‪i =1‬‬

‫‪m‬‬

‫= ‪ s0‬ﻓﻔﻲ ﻣﺜﺎﻟﻨﺎ‬

‫‪44.2 + 44.3 + 44.4 + 43.4 + 42.8 + 44.3‬‬ ‫‪= 43.9‬‬ ‫‪6‬‬

‫= ‪s0‬‬

‫وﺑﺎﻟﺘﺎﻟﻲ ﻳﻜﻮن‬ ‫‪s1 = α z1 + (1 − α ) s0 = 0.2 ( 44.2 ) + 0.8 ( 43.9 ) = 8.84 + 35.12 = 43.96‬‬ ‫‪s2 = α z2 + (1 − α ) s1 = 0.2 ( 44.3) + 0.8 ( 43.96 ) = 8.86 + 35.168 = 44.028‬‬ ‫وهﻜﺬا ﻧﺴﺘﻤﺮ ﺣﺘﻲ ﺁﺧﺮ ﻣﺸﺎهﺪة ﻓﻴﻨﺘﺞ اﻟﺘﺎﻟﻲ‪:‬‬ ‫‪RESI1‬‬

‫‪FITS1‬‬

‫‪SMOO1‬‬

‫‪Metals‬‬

‫‪Time‬‬

‫‪0.30000‬‬ ‫‪0.34000‬‬ ‫‪0.37200‬‬ ‫‪-0.70240‬‬ ‫‪-1.16192‬‬ ‫‪0.57046‬‬ ‫‪0.55637‬‬ ‫‪0.84510‬‬ ‫‪0.27608‬‬ ‫‪-1.07914‬‬ ‫‪-1.36331‬‬ ‫‪-1.29065‬‬ ‫‪-1.23252‬‬ ‫‪-1.38601‬‬ ‫‪-2.80881‬‬ ‫‪-0.34705‬‬ ‫‪0.12236‬‬ ‫‪0.79789‬‬ ‫‪-0.06169‬‬ ‫‪0.65065‬‬ ‫‪0.62052‬‬ ‫‪0.09642‬‬ ‫‪0.27713‬‬ ‫‪0.02171‬‬ ‫‪-0.28264‬‬ ‫‪-0.12611‬‬ ‫‪-0.40089‬‬ ‫‪0.27929‬‬ ‫‪0.92343‬‬ ‫‪1.83875‬‬ ‫‪1.27100‬‬ ‫‪1.51680‬‬ ‫‪1.01344‬‬ ‫‪0.91075‬‬ ‫‪1.02860‬‬

‫‪43.9000‬‬ ‫‪43.9600‬‬ ‫‪44.0280‬‬ ‫‪44.1024‬‬ ‫‪43.9619‬‬ ‫‪43.7295‬‬ ‫‪43.8436‬‬ ‫‪43.9549‬‬ ‫‪44.1239‬‬ ‫‪44.1791‬‬ ‫‪43.9633‬‬ ‫‪43.6906‬‬ ‫‪43.4325‬‬ ‫‪43.1860‬‬ ‫‪42.9088‬‬ ‫‪42.3470‬‬ ‫‪42.2776‬‬ ‫‪42.3021‬‬ ‫‪42.4617‬‬ ‫‪42.4494‬‬ ‫‪42.5795‬‬ ‫‪42.7036‬‬ ‫‪42.7229‬‬ ‫‪42.7783‬‬ ‫‪42.7826‬‬ ‫‪42.7261‬‬ ‫‪42.7009‬‬ ‫‪42.6207‬‬ ‫‪42.6766‬‬ ‫‪42.8613‬‬ ‫‪43.2290‬‬ ‫‪43.4832‬‬ ‫‪43.7866‬‬ ‫‪43.9892‬‬ ‫‪44.1714‬‬

‫‪43.9600‬‬ ‫‪44.0280‬‬ ‫‪44.1024‬‬ ‫‪43.9619‬‬ ‫‪43.7295‬‬ ‫‪43.8436‬‬ ‫‪43.9549‬‬ ‫‪44.1239‬‬ ‫‪44.1791‬‬ ‫‪43.9633‬‬ ‫‪43.6906‬‬ ‫‪43.4325‬‬ ‫‪43.1860‬‬ ‫‪42.9088‬‬ ‫‪42.3470‬‬ ‫‪42.2776‬‬ ‫‪42.3021‬‬ ‫‪42.4617‬‬ ‫‪42.4494‬‬ ‫‪42.5795‬‬ ‫‪42.7036‬‬ ‫‪42.7229‬‬ ‫‪42.7783‬‬ ‫‪42.7826‬‬ ‫‪42.7261‬‬ ‫‪42.7009‬‬ ‫‪42.6207‬‬ ‫‪42.6766‬‬ ‫‪42.8613‬‬ ‫‪43.2290‬‬ ‫‪43.4832‬‬ ‫‪43.7866‬‬ ‫‪43.9892‬‬ ‫‪44.1714‬‬ ‫‪44.3771‬‬

‫‪44.2‬‬ ‫‪44.3‬‬ ‫‪44.4‬‬ ‫‪43.4‬‬ ‫‪42.8‬‬ ‫‪44.3‬‬ ‫‪44.4‬‬ ‫‪44.8‬‬ ‫‪44.4‬‬ ‫‪43.1‬‬ ‫‪42.6‬‬ ‫‪42.4‬‬ ‫‪42.2‬‬ ‫‪41.8‬‬ ‫‪40.1‬‬ ‫‪42.0‬‬ ‫‪42.4‬‬ ‫‪43.1‬‬ ‫‪42.4‬‬ ‫‪43.1‬‬ ‫‪43.2‬‬ ‫‪42.8‬‬ ‫‪43.0‬‬ ‫‪42.8‬‬ ‫‪42.5‬‬ ‫‪42.6‬‬ ‫‪42.3‬‬ ‫‪42.9‬‬ ‫‪43.6‬‬ ‫‪44.7‬‬ ‫‪44.5‬‬ ‫‪45.0‬‬ ‫‪44.8‬‬ ‫‪44.9‬‬ ‫‪45.2‬‬

‫‪1‬‬ ‫‪2‬‬ ‫‪3‬‬ ‫‪4‬‬ ‫‪5‬‬ ‫‪6‬‬ ‫‪7‬‬ ‫‪8‬‬ ‫‪9‬‬ ‫‪10‬‬ ‫‪11‬‬ ‫‪12‬‬ ‫‪13‬‬ ‫‪14‬‬ ‫‪15‬‬ ‫‪16‬‬ ‫‪17‬‬ ‫‪18‬‬ ‫‪19‬‬ ‫‪20‬‬ ‫‪21‬‬ ‫‪22‬‬ ‫‪23‬‬ ‫‪24‬‬ ‫‪25‬‬ ‫‪26‬‬ ‫‪27‬‬ ‫‪28‬‬ ‫‪29‬‬ ‫‪30‬‬ ‫‪31‬‬ ‫‪32‬‬ ‫‪33‬‬ ‫‪34‬‬ ‫‪35‬‬

‫‪١٧٤‬‬

‫‪0.82288‬‬ ‫‪0.45830‬‬ ‫‪0.86664‬‬ ‫‪1.39331‬‬ ‫‪1.71465‬‬ ‫‪2.07172‬‬ ‫‪2.45738‬‬ ‫‪1.96590‬‬ ‫‪2.37272‬‬ ‫‪1.69818‬‬ ‫‪1.85854‬‬ ‫‪2.08683‬‬ ‫‪1.66947‬‬ ‫‪0.93557‬‬ ‫‪1.04846‬‬ ‫‪0.53877‬‬ ‫‪1.53101‬‬ ‫‪1.22481‬‬ ‫‪1.17985‬‬ ‫‪0.54388‬‬ ‫‪1.13510‬‬ ‫‪0.40808‬‬ ‫‪-0.07353‬‬ ‫‪-1.15883‬‬ ‫‪-2.02706‬‬

‫‪44.3771‬‬ ‫‪44.5417‬‬ ‫‪44.6334‬‬ ‫‪44.8067‬‬ ‫‪45.0853‬‬ ‫‪45.4283‬‬ ‫‪45.8426‬‬ ‫‪46.3341‬‬ ‫‪46.7273‬‬ ‫‪47.2018‬‬ ‫‪47.5415‬‬ ‫‪47.9132‬‬ ‫‪48.3305‬‬ ‫‪48.6644‬‬ ‫‪48.8515‬‬ ‫‪49.0612‬‬ ‫‪49.1690‬‬ ‫‪49.4752‬‬ ‫‪49.7202‬‬ ‫‪49.9561‬‬ ‫‪50.0649‬‬ ‫‪50.2919‬‬ ‫‪50.3735‬‬ ‫‪50.3588‬‬ ‫‪50.1271‬‬

‫‪44.5417‬‬ ‫‪44.6334‬‬ ‫‪44.8067‬‬ ‫‪45.0853‬‬ ‫‪45.4283‬‬ ‫‪45.8426‬‬ ‫‪46.3341‬‬ ‫‪46.7273‬‬ ‫‪47.2018‬‬ ‫‪47.5415‬‬ ‫‪47.9132‬‬ ‫‪48.3305‬‬ ‫‪48.6644‬‬ ‫‪48.8515‬‬ ‫‪49.0612‬‬ ‫‪49.1690‬‬ ‫‪49.4752‬‬ ‫‪49.7202‬‬ ‫‪49.9561‬‬ ‫‪50.0649‬‬ ‫‪50.2919‬‬ ‫‪50.3735‬‬ ‫‪50.3588‬‬ ‫‪50.1271‬‬ ‫‪49.7216‬‬

‫‪45.2‬‬ ‫‪45.0‬‬ ‫‪45.5‬‬ ‫‪46.2‬‬ ‫‪46.8‬‬ ‫‪47.5‬‬ ‫‪48.3‬‬ ‫‪48.3‬‬ ‫‪49.1‬‬ ‫‪48.9‬‬ ‫‪49.4‬‬ ‫‪50.0‬‬ ‫‪50.0‬‬ ‫‪49.6‬‬ ‫‪49.9‬‬ ‫‪49.6‬‬ ‫‪50.7‬‬ ‫‪50.7‬‬ ‫‪50.9‬‬ ‫‪50.5‬‬ ‫‪51.2‬‬ ‫‪50.7‬‬ ‫‪50.3‬‬ ‫‪49.2‬‬ ‫‪48.1‬‬

‫‪36‬‬ ‫‪37‬‬ ‫‪38‬‬ ‫‪39‬‬ ‫‪40‬‬ ‫‪41‬‬ ‫‪42‬‬ ‫‪43‬‬ ‫‪44‬‬ ‫‪45‬‬ ‫‪46‬‬ ‫‪47‬‬ ‫‪48‬‬ ‫‪49‬‬ ‫‪50‬‬ ‫‪51‬‬ ‫‪52‬‬ ‫‪53‬‬ ‫‪54‬‬ ‫‪55‬‬ ‫‪56‬‬ ‫‪57‬‬ ‫‪58‬‬ ‫‪59‬‬ ‫‪60‬‬

‫اﻟﻌﻤﻮد اﻟﺮاﺑﻊ ‪ SMOO١‬ﻳﺤﻮى اﻟﻘﻴﻢ اﻟﻤﻤﻬﺪة أي ‪ si , i = 1, 2,...,60‬اﻟﻌﻤﻮد اﻟﺨﺎﻣﺲ ‪FITS١‬‬ ‫ﻳﺤﻮى اﻟﻘﻴﻢ اﻟﻤﻄﺒﻘﺔ أي ‪ zˆi = si −1 , i = 1, 2,...,60‬اﻟﻌﻤﻮد اﻟﺨﺎﻣﺲ ‪ RESI١‬ﻳﺤﻮي اﻷﺧﻄﺎء‬ ‫)اﻟﺒﻮاﻗﻲ ‪ (Residuals‬أي ‪ei = zi − zˆi , i = 1, 2,...,60‬‬ ‫‪ -٢‬ﻳﺆﺧﺬ آﻤﺘﻨﺒﺊ ﻟﻠﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﺁﺧﺮ ﻗﻴﻤﺔ ﻣﻤﻬﺪة أي‪:‬‬ ‫‪z n ( l ) = sn , l > 0‬‬ ‫ﻓﻔﻲ اﻟﻤﺜﺎل اﻟﺤﺎﻟﻲ‬ ‫‪z60 ( l ) = 49.7216, l > 0‬‬ ‫ﺗﺆﺧﺬ اﻟﺘﻨﺒﺆات ﻟﻠﻘﻴﻢ اﻟـ ‪ ٦‬اﻟﻤﺴﺘﻘﺒﻠﻴﺔ أي ﻟﻠﻘﻴﻢ ‪ zn +1 , zn +2 ,..., zn +6‬أو ﻓﻲ هﺬا اﻟﻤﺜﺎل‬ ‫‪ z61 , z62 ,..., z66‬آﺎﻟﺘﺎﻟﻲ‪:‬‬ ‫‪z60 (1) = z60 ( 2 ) = L = z60 ( 6 ) = 49.7216‬‬

‫‪ -٣‬ﻟﺤﺴﺎب ﻓﺘﺮات ﺗﻨﺒﺆ ‪ 95%‬ﻧﺤﺴﺐ اﻟﻜﻤﻴﺎت ‪ ⎡⎣ zn ( l ) ± 1.96σˆ ⎤⎦ , l > 0‬أي‬ ‫] ˆ‪ [ 49.7216 ± 1.96σ‬ﻟﺠﻤﻴﻊ ﻗﻴﻢ اﻟﺘﻨﺒﺆات اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ ،‬ﻧﺄﺧﺬ اﻟﻘﻴﻤﺔ ‪ MSD = 1.45392‬آﻤﻘﺪر‬

‫ﻟـ ‪ σ 2‬أي ‪ σˆ 2 = 1.45392‬ﻓﻴﻜﻮن ‪ σˆ = 1.205786‬وﻋﻠﻴﻪ ﺗﻜﻮن ﻓﺘﺮة ﺗﻨﺒﺆ ‪ 95%‬ﻟﺠﻤﻴﻊ‬ ‫اﻟﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ هﻲ‪:‬‬ ‫]‪⎡⎣ 49.7216 ± 1.96 (1.205786 ) ⎤⎦ = [ 49.7216 ± 2.3633] = [ 47.35826,52.08494‬‬ ‫أي‪:‬‬ ‫‪z60+l ∈ [ 47.3582,50.0849] , l > 0 with probability 0.95‬‬ ‫ﻣﻼﺣﻈﺔ‪ :‬ﺗﺤﺴﺐ ‪ MSD‬آﺎﻵﺗﻲ‬

‫‪١٧٥‬‬

‫) ‪− zˆi‬‬

‫‪n‬‬

‫‪∑( z‬‬

‫‪i‬‬

‫‪n −1‬‬

‫‪i =1‬‬

‫= ˆ‪MSD = σ‬‬ ‫‪2‬‬

‫ﺗﻤﺮﻳﻦ‪:‬‬ ‫ﻃﺒﻖ ﺗﻤﻬﻴﺪ اﺳﻲ ﺑﺴﻴﻂ ﻋﻠﻲ اﻟﻤﺸﺎهﺪات اﻟﺴﺎﺑﻘﺔ ﻣﺴﺘﺨﺪﻣﺎ ‪ α = 0.3,0.4,0.5‬وﻗﺮر اﻳﻬﺎ اﻓﻀﻞ‬ ‫ﻟﻠﺘﻨﺒﺆ ﻋﻦ اﻟﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ؟‪.‬‬

‫‪١٧٦‬‬

‫اﻟﻔﺼﻞ اﻟﺜﺎﻧﻲ ﻋﺸﺮ‬ ‫اﻟﺘﻤﻬﻴﺪ و اﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ اﻟﺘﻤﻬﻴﺪ اﻻﺳﻲ اﻟﻤﺰدوج ‪Using Double‬‬ ‫‪: Exponential Smoothing for Forecasting‬‬ ‫أوﻻ‪ :‬ﻃﺮﻳﻘﺔ ﺑﺮاون ‪:Brown’s Method‬‬

‫ﻟﻤﺸﺎهﺪات ‪ z1 , z2 ,K , zn −1 , zn −2‬وﻟﺜﺎﺑﺖ ﺗﻤﻬﻴﺪ ‪ 0 < α < 1‬ﻧﻮﺟﺪ اﻟﺘﺎﻟﻲ‪:‬‬

‫‪s = α zt + (1 − α ) st(−1) , t = 1,2,..., n‬‬ ‫)‪(1‬‬ ‫‪t‬‬

‫‪1‬‬

‫ﺣﻴﺚ ) (‪ st‬ﺗﻤﻬﻴﺪ اﺳﻲ ﺑﺴﻴﻂ ‪) st( ) = st‬اﻧﻈﺮ اﻟﺘﻤﻬﻴﺪ اﻻﺳﻲ اﻟﺒﺴﻴﻂ( و )‪ (1‬ﺗﺮﻣﺰ اﻟﻲ درﺟﺔ هﺬا‬ ‫اﻟﺘﻤﻬﻴﺪ‬ ‫)‪(2‬‬ ‫)‪(1‬‬ ‫)‪(2‬‬ ‫‪st = α st + (1 − α ) st −1 , t = 1,2,..., n‬‬ ‫‪1‬‬

‫‪1‬‬

‫ﺣﻴﺚ ) (‪ st‬ﺗﻤﻬﻴﺪ اﺳﻲ ﻣﺰدوج و ) ‪ ( 2‬ﺗﺮﻣﺰ اﻟﻲ درﺟﺔ هﺬا اﻟﺘﻤﻬﻴﺪ‬ ‫‪2‬‬

‫‪at = 2 st( ) − st( ) , t = 1, 2,..., n‬‬ ‫‪1‬‬

‫‪2‬‬

‫‪α‬‬ ‫‪1‬‬ ‫‪2‬‬ ‫‪st( ) − st( ) , t = 1, 2,..., n‬‬ ‫‪1−α‬‬

‫)‬

‫ﺗﺤﺴﺐ اﻟﻘﻴﻢ اﻟﻤﻄﺒﻘﺔ ﻣﻦ اﻟﻤﻌﺎدﻟﺔ‬

‫‪t = 1, 2,..., n‬‬

‫وﺗﺤﺴﺐ اﻟﺘﻨﺒﺆات ﻟﻠﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ‪ zn +l , l > 0‬ﻣﻦ‬

‫‪zˆt = at + bt t ,‬‬

‫‪zn ( l ) = an + bn l, l > 0‬‬

‫ﺣﺴﺎب اﻟﻘﻴﻢ اﻻوﻟﻴﺔ ) (‪ s0‬و ) (‪ : s0‬ﻣﻦ اﻟﻌﻼﻗﺎت اﻟﺴﺎﺑﻘﺔ ﻧﺠﺪ‬ ‫‪1‬‬

‫(‬

‫= ‪bt‬‬

‫‪2‬‬

‫‪b0‬‬ ‫‪b0‬‬ ‫ﻧﻮﺟﺪ ‪ a0‬و ‪ b0‬ﺑﺈﻧﺤﺪار اﻟﻤﺸﺎهﺪات ﻋﻠﻲ اﻟﺰﻣﻦ ‪t = 1, 2,..., n‬‬ ‫ˆ‪ a0 = α‬و ˆ‪b0 = β‬‬

‫‪1−α‬‬

‫‪α‬‬

‫‪1−α‬‬

‫‪α‬‬

‫‪s0( ) = a0 −‬‬ ‫‪1‬‬

‫‪s0( ) = a0 − 2‬‬ ‫‪2‬‬

‫‪ zt = α + β t + et ,‬وﻳﻜﻮن‬

‫ﺛﺎﻧﻴﺎ‪ :‬ﻃﺮﻳﻘﺔ هﻮﻟﺖ ‪:Holt’s Method‬‬

‫ﻟﻤﺸﺎهﺪات ‪ z1 , z2 ,K , zn −1 , zn −2‬وﻟﺜﺎﺑﺘﻲ ﺗﻤﻬﻴﺪ ‪ 0 < α < 1‬و ‪ 0 < γ < 1‬ﻧﻮﺟﺪ اﻟﺘﺎﻟﻲ‪:‬‬ ‫‪st = α zt + (1 − α )( st −1 + bt −1 ) , t = 1, 2,..., n‬‬ ‫‪t = 1, 2,..., n‬‬ ‫ﻧﺤﺴﺐ اﻟﻘﻴﻢ اﻟﻤﻄﺒﻘﺔ ﻣﻦ‬

‫‪bt = γ ( st − st −1 ) + (1 − γ ) bt −1 ,‬‬ ‫‪t = 1, 2,..., n‬‬

‫واﻟﺘﻨﺒﺆات ﻟﻠﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﻣﻦ‬

‫‪zˆt = st + bt t ,‬‬

‫‪zn ( l ) = sn + bn l, l > 0‬‬

‫ﻧﺤﺴﺐ اﻟﻘﻴﻢ اﻻوﻟﻴﺔ ‪ s0‬و ‪ b0‬ﻣﻦ‬ ‫‪١٧٧‬‬

s0 = z1 b0 = z2 − z1

b0 =

or

( z2 − z1 ) + ( z3 − z2 ) = ( z3 − z1 )

or 2 2 ( z − z ) + ( z3 − z2 ) + ( z4 − z3 ) ( z4 − z1 ) b0 = 2 1 = 3 3 :‫ﻣﺜﺎل‬ EMPLOY.MTW ‫ﺗﺤﻤﻞ اﻟﺒﻴﺎﻧﺎت ﻣﻦ ورﻗﺔ اﻟﻌﻤﻞ‬ MTB > Retrieve 'E:\Mtbwin\DATA\EMPLOY.MTW'.

Metals ‫ﺳﻮف ﻧﺴﺘﺨﺪم اﻟﻤﺸﺎهﺪات ﻓﻲ اﻟﺘﻐﻴﺮ‬ Metals 44.2 43.1 42.4 42.9 45.0 49.4 50.5

44.3 42.6 43.1 43.6 45.5 50.0 51.2

44.4 42.4 43.2 44.7 46.2 50.0 50.7

43.4 42.2 42.8 44.5 46.8 49.6 50.3

42.8 41.8 43.0 45.0 47.5 49.9 49.2

44.3 40.1 42.8 44.8 48.3 49.6 48.1

44.4 42.0 42.5 44.9 48.3 50.7

44.8 42.4 42.6 45.2 49.1 50.7

44.4 43.1 42.3 45.2 48.9 50.9

:‫ﻧﺮﺳﻢ هﺬﻩ اﻟﻤﺸﺎهﺪات‬ MTB > TSPlot 'Metals'; SUBC> Index; SUBC> TDisplay 11; SUBC> Symbol; SUBC> Connect.

Metals

50

45

40 Index

10

20

30

١٧٨

40

50

60

‫ ﻧﺴﺘﺨﺪم اﻵن‬،‫ﻧﻄﺒﻖ اﻵن ﺗﻤﻬﻴﺪا ﻟﻬﺬﻩ اﻟﻤﺸﺎهﺪات ﺑﺈﺳﺘﺨﺪام اﻟﺘﻤﻬﻴﺪ اﻻﺳﻲ اﻟﻤﺰدوج ﺑﻄﺮﻳﻘﺔ ﺑﺮاون‬ ‫ ﺑﺄوزان ﻣﺘﺴﺎوﻳﺔ ﺳﻮف ﻧﺄﺧﺬهﺎ‬WEIGHT ‫ ﻣﻊ اﻻﻣﺮ اﻟﻔﺮﻋﻲ‬%DES (Macro) ‫اﻟﺒﺮﻣﺞ‬ 0.2 MTB > %DES 'Metals'; SUBC> Weight 0.2 0.2; SUBC> Forecasts 6; SUBC> Title "Brown's Double Exponential Smoothing"; SUBC> Table.

Double Exponential Smoothing Data Length NMissing

Metals 60.0000 0

Smoothing Constants Alpha (level): 0.2 Gamma (trend): 0.2 Accuracy Measures MAPE: 2.16187 MAD: 0.97032 MSD: 1.62936

Time Metals 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

44.2 44.3 44.4 43.4 42.8 44.3 44.4 44.8 44.4 43.1 42.6 42.4 42.2 41.8 40.1 42.0 42.4 43.1 42.4 43.1 43.2 42.8 43.0 42.8 42.5 42.6 42.3 42.9 43.6 44.7

Smooth

Predict

Error

41.7739 42.4976 43.1686 43.5546 43.7373 44.1459 44.4990 44.8575 45.0620 44.9391 44.6673 44.3271 43.9378 43.4769 42.7011 42.3565 42.1465 42.1286 42.0132 42.0763 42.1877 42.2374 42.3396 42.4078 42.4181 42.4495 42.4207 42.5129 42.7420 43.1797

41.1674 42.0470 42.8607 43.5933 43.9716 44.1074 44.5238 44.8719 45.2275 45.3989 45.1841 44.8088 44.3723 43.8962 43.3514 42.4456 42.0831 41.8857 41.9164 41.8204 41.9347 42.0967 42.1745 42.3098 42.3976 42.4119 42.4509 42.4161 42.5276 42.7996

3.03257 2.25303 1.53927 -0.19330 -1.17163 0.19257 -0.12377 -0.07189 -0.82751 -2.29891 -2.58407 -2.40884 -2.17229 -2.09617 -3.25142 -0.44557 0.31694 1.21426 0.48355 1.27964 1.26533 0.70327 0.82549 0.49024 0.10244 0.18809 -0.15090 0.48394 1.07245 1.90036

١٧٩

31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

44.5 45.0 44.8 44.9 45.2 45.2 45.0 45.5 46.2 46.8 47.5 48.3 48.3 49.1 48.9 49.4 50.0 50.0 49.6 49.9 49.6 50.7 50.7 50.9 50.5 51.2 50.7 50.3 49.2 48.1

43.5507 43.9854 44.3338 44.6511 44.9749 45.2430 45.4157 45.6373 45.9491 46.3285 46.7909 47.3492 47.8339 48.4002 48.8413 49.2966 49.7849 50.1841 50.4162 50.6292 50.7104 50.9509 51.1334 51.3019 51.3407 51.4782 51.4770 51.3649 51.0127 50.4384

43.3133 43.7317 44.2172 44.5889 44.9187 45.2538 45.5197 45.6716 45.8863 46.2106 46.6136 47.1115 47.7173 48.2253 48.8267 49.2707 49.7311 50.2302 50.6202 50.8114 50.9880 51.0137 51.2417 51.4024 51.5509 51.5477 51.6712 51.6311 51.4659 51.0230

1.18668 1.26826 0.58280 0.31112 0.28133 -0.05376 -0.51967 -0.17162 0.31368 0.58938 0.88637 1.18850 0.58266 0.87469 0.07332 0.12930 0.26890 -0.23017 -1.02022 -0.91145 -1.38797 -0.31368 -0.54169 -0.50244 -1.05093 -0.34770 -0.97120 -1.33115 -2.26587 -2.92300

Row

Period

Forecast

Lower

Upper

1 2 3 4 5 6

61 62 63 64 65 66

50.3318 50.2252 50.1186 50.0120 49.9054 49.7988

47.9545 47.7984 47.6384 47.4749 47.3080 47.1381

52.7091 52.6520 52.5987 52.5490 52.5027 52.4594

Brown's Double Exponential Smoothing

Actual Predicted Forecast

Metals

50

Actual Predicted Forecast

45

Smoothing Constants Alpha (level): 0.200 Gamma (trend):0.200 MAPE: MAD: MSD:

40 0

10

20

30

Time

١٨٠

40

50

60

2.16187 0.97032 1.62936

: b0 ‫ و‬a0 ‫إﻳﺠﺎد‬ MTB > DATA> DATA> MTB >

set c4 1:60 end regr c3 1 c4

Regression Analysis The regression equation is Metals = 41.0 + 0.152 C4

1−α

‫ وﻣﻨﻬﺎ ﻧﺤﺴﺐ‬b0 = 0.152 ‫ و‬a0 = 41.0 ‫إذا‬

0.8 ( 0.152 ) = 41.608 α 0.2 1−α ⎛ 0.8 ⎞ 2 s0( ) = a0 − 2 b0 = 41.0 − 2 ⎜ ⎟ ( 0.152 ) = 42.216 α ⎝ 0.2 ⎠ 1 s1( ) = ( 0.2 )( 44.2 ) + ( 0.8 )( 41.608 ) = 42.1264 s0( ) = a0 − 1

b0 = 41.0 −

s1( ) = ( 0.2 )( 42.1264 ) + ( 0.8 )( 42.216 ) = 42.19808 2

a1 = ( 2 )( 42.1264 ) − 42.19808 = 42.05472

( 0.2 ) ( 42.19808 − 42.05472 ) = 0.03584 ( 0.8) zˆ1 = 42.05472 + ( 0.03584 )(1) = 42.09056

b1 =

… ‫وهﻜﺬا اﻟﺦ‬ .‫ آﻤﺎ ﻓﻲ اﻷﻣﺜﻠﺔ اﻟﺴﺎﺑﻘﺔ‬MSD ‫ﺗﺤﺴﺐ ﻓﺘﺮات اﻟﺘﻨﺒﺆ ﺑﺈﺳﺘﺨﺪام‬ :‫ﻣﺜﺎل‬ WEIGHT ‫ ﻣﻊ اﻻﻣﺮ اﻟﻔﺮﻋﻲ‬%DES (Macro) ‫ﻟﺘﻄﺒﻴﻖ ﻃﺮﻳﻘﺔ هﻮﻟﺖ ﻧﺴﺘﺨﺪم اﻵن اﻟﺒﺮﻣﺞ‬ γ = 0.3 ‫ و‬α = 0.2 ‫ﺑﺄوزان ﻣﺨﺘﻠﻔﺔ ﺳﻮف ﻧﺄﺧﺬ‬ MTB > RETR 'E:\Mtbwin\DATA\EMPLOY.MTW'. Retrieving worksheet from file: E:\Mtbwin\DATA\EMPLOY.MTW Worksheet was saved on 6/ 5/1996 MTB > %DES 'Metals'; SUBC> Weight 0.2 0.3; SUBC> Forecasts 6; SUBC> Title "Holt's Double Exponential Smoothing"; SUBC> Table.

Double Exponential Smoothing Data Length NMissing

Metals 60.0000 0

١٨١

Smoothing Constants Alpha (level): 0.2 Gamma (trend): 0.3 Accuracy Measures MAPE: 2.15656 MAD: 0.96328 MSD: 1.56274

Time

Metals

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48

44.2 44.3 44.4 43.4 42.8 44.3 44.4 44.8 44.4 43.1 42.6 42.4 42.2 41.8 40.1 42.0 42.4 43.1 42.4 43.1 43.2 42.8 43.0 42.8 42.5 42.6 42.3 42.9 43.6 44.7 44.5 45.0 44.8 44.9 45.2 45.2 45.0 45.5 46.2 46.8 47.5 48.3 48.3 49.1 48.9 49.4 50.0 50.0

Smooth 41.7739 42.5461 43.2891 43.7501 43.9779 44.3895 44.7334 45.0685 45.2405 45.0676 44.7113 44.2595 43.7466 43.1634 42.2751 41.8139 41.5361 41.5057 41.4371 41.5799 41.8054 41.9895 42.2254 42.4205 42.5395 42.6522 42.6793 42.7982 43.0394 43.4861 43.8762 44.3257 44.6858 45.0007 45.3066 45.5449 45.6749 45.8384 46.0888 46.4159 46.8406 47.3799 47.8665 48.4419 48.9016 49.3693 49.8653 50.2702

Predict 41.1674 42.1076 43.0113 43.8376 44.2724 44.4118 44.8167 45.1356 45.4506 45.5595 45.2391 44.7244 44.1332 43.5043 42.8188 41.7674 41.3202 41.1072 41.1964 41.1999 41.4568 41.7869 42.0317 42.3257 42.5493 42.6653 42.7741 42.7728 42.8993 43.1826 43.7202 44.1571 44.6573 45.0259 45.3333 45.6312 45.8436 45.9230 46.0610 46.3199 46.6757 47.1499 47.7582 48.2774 48.9020 49.3617 49.8317 50.3378

Error 3.03257 2.19238 1.38868 -0.43760 -1.47237 -0.11184 -0.41671 -0.33560 -1.05057 -2.45952 -2.63911 -2.32443 -1.93322 -1.70426 -2.71884 0.23263 1.07985 1.99283 1.20365 1.90008 1.74322 1.01314 0.96829 0.47431 -0.04933 -0.06528 -0.47413 0.12724 0.70070 1.51743 0.77977 0.84285 0.14275 -0.12590 -0.13327 -0.43116 -0.84361 -0.42295 0.13895 0.48014 0.82428 1.15014 0.54181 0.82265 -0.00205 0.03832 0.16832 -0.33779

١٨٢

49 50 51 52 53 54 55 56 57 58 59 60

49.6 49.9 49.6 50.7 50.7 50.9 50.5 51.2 50.7 50.3 49.2 48.1

50.4979 50.6862 50.7296 50.9166 51.0532 51.1813 51.1869 51.2901 51.2673 51.1350 50.7591 50.1448

50.7224 50.8827 51.0121 50.9708 51.1415 51.2516 51.3586 51.3127 51.4092 51.3438 51.1489 50.6560

-1.12240 -0.98275 -1.41206 -0.27079 -0.44152 -0.35162 -0.85860 -0.11267 -0.70916 -1.04381 -1.94889 -2.55603

Row

Period

Forecast

Lower

Upper

1 2 3 4 5 6

61 62 63 64 65 66

49.8884 49.6319 49.3755 49.1190 48.8626 48.6061

47.5283 47.1597 46.7803 46.3915 45.9946 45.5908

52.2484 52.1041 51.9707 51.8466 51.7306 51.6215

Holt's Double Exponential Smoothing

Actual Predicted Forecast

Metals

50

Actual Predicted Forecast

45

Smoothing Constants Alpha (level): 0.200 Gamma (trend):0.300 MAPE: MAD: MSD:

40 0

10

20

30

40

50

2.15656 0.96328 1.56274

60

Time

‫ ﻳﺘﻮﻗﻊ ﻋﺪم ﺗﻄﺎﺑﻖ‬:‫ ﺗﺤﻘﻖ ﻣﻦ ﺻﺤﺔ اﻟﺤﺴﺎﺑﺎت اﻟﺴﺎﺑﻘﺔ ﺑﺘﺘﺒﻊ ﺑﻌﺾ اﻟﻘﻴﻢ ﻳﺪوﻳﺎ ) ﻣﻼﺣﻈﺔ‬:‫ﺗﻤﺮﻳﻦ‬ ‫اﻟﺤﺴﺎﺑﺎت ﺗﻤﺎﻣﺎ وذﻟﻚ ﻹﺧﺘﻼف ﻃﺮﻳﻖ ﺗﻤﺜﻴﻞ اﻷﻋﺪاد ﺑﻴﻦ اﻟﺤﺎﺳﺐ واﻵﻟﺔ اﻟﺤﺎﺳﺒﺔ وآﺬﻟﻚ ﻓﻲ‬ (‫ﺗﺨﺰﻳﻦ اﻷرﻗﺎم ﻓﻲ ذاآﺮات آﻞ ﻣﻨﻬﻤﺎ‬

١٨٣

‫اﻟﻔﺼﻞ اﻟﺜﺎﻟﺚ ﻋﺸﺮ‬ ‫اﻟﺘﻤﻬﻴﺪ اﻻﺳﻲ اﻟﺜﻼﺛﻲ و اﻟﺘﻨﺒﺆ ﺑﻮاﺳﻄﺔ ﻃﺮﻳﻘﺔ وﻧﺘﺮز ﻟﻠﻤﺘﺴﻠﺴﻼت‬ ‫اﻟﻤﻮﺳﻤﻴﺔ اﻟﻤﻨﺠﺮﻓﺔ‬ ‫‪Triple Exponential Smoothing: Winters' Three‬‬‫‪Parameter Trend and Seasonality Smoothing Method‬‬ ‫ﺟﻤﻴﻊ اﻟﻄﺮق اﻟﺴﺎﺑﻘﺔ اﻟﺘﻲ درﺳﻨﺎهﺎ ﻓﻲ هﺬا اﻟﻔﺼﻞ ﻻ ﺗﻨﻔﻊ ﻟﺘﺤﻠﻴﻞ اﻟﻈﻮاهﺮ اﻟﻤﻮﺳﻤﻴﺔ ﻣﺎﻋﺪى‬ ‫ﻃﺮﻳﻘﺔ اﻟﺘﻔﻜﻴﻚ ‪ Decomposition Method‬وﻃﺮﻳﻘﺔ وﻧﺘﺮز ‪Winters' trend and‬‬ ‫‪ seasonal smoothing‬اﻟﺘﻲ ﺳﻮف ﻧﺴﺘﻌﺮﺿﻬﺎ هﻨﺎ‬

‫ﻃﺮﻳﻘﺔ وﻧﺘﺮز ﻟﻠﻤﺘﺴﻠﺴﻼت اﻟﻤﻮﺳﻤﻴﺔ اﻟﻤﻨﺠﺮﻓﺔ‬ ‫أوﻻ‪ :‬ﺗﻤﻬﺪ اﻟﻤﺸﺎهﺪات ﺗﻤﻬﻴﺪا آﻠﻴﺎ ﺑﺎﻟﻌﻼﻗﺔ‬ ‫‪zt‬‬ ‫‪+ (1 − α )( st −1 + bt −1 ) , t = 1, 2,..., n‬‬ ‫‪St − s‬‬ ‫ﺛﺎﻧﻴﺎ‪ :‬ﺗﻤﻬﻴﺪ اﻹﻧﺠﺮاف‬

‫‪st = α‬‬

‫‪bt = γ ( st − st −1 ) + (1 − γ ) bt −1 , t = 1, 2,..., n‬‬

‫ﺛﺎﻟﺜﺎ‪ :‬ﺗﻤﻬﻴﺪ اﻟﻤﻮﺳﻤﻴﺔ‬ ‫‪zt‬‬ ‫‪+ (1 − β ) St − s , t = 1, 2,..., n‬‬ ‫‪st‬‬ ‫ﺣﻴﺚ ‪ Si‬هﻲ اﻟﻤﺮآﺒﺔ اﻟﻤﻮﺳﻤﻴﺔ ﻋﻨﺪ اﻟﺰﻣﻦ ‪ i‬و ‪ s‬هﻲ دورة اﻟﻤﻮﺳﻤﻴﺔ‬ ‫اﻟﻘﻴﻢ اﻟﻤﻄﺒﻘﺔ ﺗﻌﻄﻲ ﺑﺎﻟﻌﻼﻗﺔ‬ ‫‪zˆt = ( st + bt t ) St − s , t = 1, 2,..., n‬‬ ‫واﻟﺘﻨﺒﺆات ﻣﻦ اﻟﻌﻼﻗﺔ‬ ‫‪zn ( l ) = ( sn + bn l ) Sn − s+l , l > 0‬‬

‫‪St = β‬‬

‫ﻣﻦ اﻟﺼﻌﺐ ﺟﺪا ﺗﺘﺒﻊ ﻃﺮﻳﻘﺔ وﻧﺘﺮز ﺑﺎﻟﺤﺴﺎﺑﺎت اﻟﻴﺪوﻳﺔ ﺣﻴﺚ ان اﻟﻘﻴﻢ اﻻوﻟﻴﺔ ﺗﺤﺴﺐ ﺑﺨﻮارزﻣﺎت‬ ‫ﻏﻴﺮ ﺧﻄﻴﺔ ﺑﺈﺳﺘﺨﺪام اﻟﺤﺎﺳﺐ وﻟﻬﺬا ﻟﻦ ﻧﺴﺘﻌﺮﺿﻬﺎ هﻨﺎ وﻧﻜﺘﻔﻲ ﺑﺎﻟﻨﺘﺎﺋﺞ اﻟﻤﺨﺮﺟﺔ ﻣﻦ اﻟﺤﺎﺳﺐ‪.‬‬ ‫ﻣﺜﺎل‪:‬‬ ‫ﺗﺤﻤﻞ اﻟﺒﻴﺎﻧﺎت ﻣﻦ ورﻗﺔ اﻟﻌﻤﻞ ‪EMPLOY.MTW‬‬ ‫‪'E:\Mtbwin\DATA\EMPLOY.MTW'.‬‬

‫‪MTB > Retrieve‬‬

‫ﺳﻮف ﻧﺴﺘﺨﺪم اﻟﻤﺸﺎهﺪات ﻓﻲ اﻟﺘﻐﻴﺮ ‪Food‬‬ ‫‪Food‬‬

‫‪52.1‬‬ ‫‪51.5‬‬ ‫‪53.5‬‬ ‫‪54.3‬‬

‫‪53.4‬‬ ‫‪52.3‬‬ ‫‪53.1‬‬ ‫‪54.6‬‬

‫‪55.3‬‬ ‫‪53.6‬‬ ‫‪53.3‬‬ ‫‪54.2‬‬

‫‪58.2‬‬ ‫‪55.3‬‬ ‫‪53.9‬‬ ‫‪54.8‬‬

‫‪66.9‬‬ ‫‪58.5‬‬ ‫‪55.6‬‬ ‫‪55.8‬‬

‫‪70.7‬‬ ‫‪69.3‬‬ ‫‪60.1‬‬ ‫‪57.9‬‬ ‫‪57.7‬‬

‫‪65.3‬‬ ‫‪69.6‬‬ ‫‪68.9‬‬ ‫‪62.6‬‬ ‫‪60.5‬‬

‫‪١٨٤‬‬

‫‪56.5‬‬ ‫‪64.2‬‬ ‫‪68.8‬‬ ‫‪70.3‬‬ ‫‪66.4‬‬

‫‪53.4‬‬ ‫‪55.5‬‬ ‫‪63.6‬‬ ‫‪69.4‬‬ ‫‪75.5‬‬

‫‪52.5‬‬ ‫‪53.3‬‬ ‫‪57.1‬‬ ‫‪64.7‬‬ ‫‪73.3‬‬

‫‪53.2‬‬ ‫‪52.4‬‬ ‫‪52.2‬‬ ‫‪57.1‬‬ ‫‪68.1‬‬

‫‪53.0‬‬ ‫‪51.5‬‬ ‫‪51.5‬‬ ‫‪53.9‬‬ ‫‪58.1‬‬

‫‪53.5‬‬ ‫‪51.5‬‬ ‫‪51.7‬‬ ‫‪53.5‬‬ ‫‪54.8‬‬

‫وﻧﺮﺳﻢ اﻟﻤﺸﺎهﺪات‬

75

Food

70 65 60 55 50 Index

10

20

30

40

50

60

:‫ ﻧﻄﺒﻖ اﻵن ﻃﺮﻳﻘﺔ وﻧﺘﺮز آﺎﻟﺘﺎﻟﻲ‬١٢ ‫ﻧﻼﺣﻆ ان اﻟﻈﺎهﺮة ﻣﻮﺳﻤﻴﺔ ﺑﺪورة‬ Additive Model ‫ ﻟﻠﻨﻤﻮذج اﻹﺿﺎﻓﻲ‬:‫أوﻻ‬ zt = bt + St + et , t = 1, 2,..., n MTB > %Wintadd 'Food' 12; SUBC> Weight 0.2 0.2 0.2; SUBC> Forecasts 12; SUBC> Title "Wintrs' Trend and Seasonal Smoothing"; SUBC> Table.

Winters' additive model Data Length NMissing

Food 60.0000 0

Smoothing Constants Alpha (level): 0.2 Gamma (trend): 0.2 Delta (seasonal): 0.2 Accuracy Measures MAPE: 1.94769 MAD: 1.15100 MSD: 2.66711

١٨٥

Time 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59

Food 53.5 53.0 53.2 52.5 53.4 56.5 65.3 70.7 66.9 58.2 55.3 53.4 52.1 51.5 51.5 52.4 53.3 55.5 64.2 69.6 69.3 58.5 55.3 53.6 52.3 51.5 51.7 51.5 52.2 57.1 63.6 68.8 68.9 60.1 55.6 53.9 53.3 53.1 53.5 53.5 53.9 57.1 64.7 69.4 70.3 62.6 57.9 55.8 54.8 54.2 54.6 54.3 54.8 58.1 68.1 73.3 75.5 66.4 60.5

Smooth 48.7755 49.6020 51.0736 52.0733 53.5117 57.4851 66.2299 71.7852 71.8932 62.3206 57.5208 55.1544 55.0393 53.6493 53.1185 52.2661 52.6528 55.7616 63.8483 68.8787 68.0386 59.2825 55.0979 53.0432 53.0377 52.1394 51.9889 51.7214 52.1875 55.0750 63.7515 68.8427 68.0160 58.9061 55.3220 53.4419 53.3084 52.6717 52.9095 52.9745 53.7952 57.2065 65.2747 70.4039 69.6200 60.5655 57.0392 55.3721 55.2403 54.6681 54.8138 54.7366 55.3001 58.5310 66.4168 71.8806 71.8794 63.7240 60.3141

Predict 49.4303 50.4197 51.9944 53.0424 54.4591 58.3901 67.0593 72.5443 72.5785 62.7787 57.7958 55.3296 55.1373 53.6258 53.0100 52.0971 52.4960 55.6369 63.7181 68.7678 67.9610 59.2585 55.0435 52.9991 53.0177 52.0907 51.9165 51.6403 52.1009 54.9923 63.7531 68.8382 68.0099 58.9356 55.3981 53.5262 53.4076 52.7666 53.0177 53.1020 53.9386 57.3484 65.4066 70.5076 69.6794 60.6497 57.2014 55.5623 55.4399 54.8422 54.9622 54.8705 55.4112 58.6176 66.4827 72.0112 72.0616 64.0437 60.7281

Error 4.06965 2.58027 1.20556 -0.54244 -1.05914 -1.89013 -1.75932 -1.84430 -5.67851 -4.57874 -2.49577 -1.92957 -3.03734 -2.12584 -1.50996 0.30287 0.80401 -0.13695 0.48187 0.83218 1.33902 -0.75851 0.25650 0.60092 -0.71765 -0.59067 -0.21651 -0.14031 0.09913 2.10774 -0.15314 -0.03822 0.89007 1.16436 0.20190 0.37384 -0.10757 0.33345 0.48233 0.39803 -0.03858 -0.24838 -0.70661 -1.10758 0.62065 1.95031 0.69858 0.23773 -0.63993 -0.64220 -0.36218 -0.57048 -0.61119 -0.51765 1.61731 1.28878 3.43843 2.35629 -0.22810

١٨٦

60

57.7

58.6397

59.0446

-1.34455

Row

Period

Forecast

Lower

Upper

1 2 3 4 5 6 7 8 9 10 11 12

61 62 63 64 65 66 67 68 69 70 71 72

58.6167 58.3236 58.8195 58.9840 59.8723 63.4804 72.0757 77.4486 77.7540 68.9067 64.6434 62.7731

55.7968 55.4449 55.8775 55.9746 56.7913 60.3243 68.8410 74.1321 74.3528 65.4180 61.0647 59.1020

61.4366 61.2023 61.7614 61.9935 62.9532 66.6365 75.3104 80.7651 81.1552 72.3954 68.2221 66.4441

Wintrs' Trend and Seasonal Smoothing

Actual

80

Predicted Forecast Actual Predicted Forecast

Food

70

60

Smoothing Constants Alpha (level): 0.200 Gamma (trend):0.200 Delta (season):0.200

50

MAPE: MAD: MSD:

0

10

20

30

40

50

60

1.94769 1.15100 2.66711

70

Time

zt = bt St + et , t = 1, 2,..., n

Multiplicative Model ‫ ﻟﻠﻨﻤﻮذج اﻟﺘﻀﺎﻋﻔﻲ‬:‫ﺛﺎﻧﻴﺎ‬

MTB > %Wintmult 'Food' 12; SUBC> Weight 0.2 0.2 0.2; SUBC> Forecasts 12; SUBC> Title "Winters' Trend and Seasonal Smoothing"; SUBC> Table.

Winters' multiplicative model Data

Food

١٨٧

Length NMissing

60.0000 0

Smoothing Constants Alpha (level): 0.2 Gamma (trend): 0.2 Delta (seasonal): 0.2 Accuracy Measures MAPE: 1.88377 MAD: 1.12068 MSD: 2.86696

Time 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44

Food 53.5 53.0 53.2 52.5 53.4 56.5 65.3 70.7 66.9 58.2 55.3 53.4 52.1 51.5 51.5 52.4 53.3 55.5 64.2 69.6 69.3 58.5 55.3 53.6 52.3 51.5 51.7 51.5 52.2 57.1 63.6 68.8 68.9 60.1 55.6 53.9 53.3 53.1 53.5 53.5 53.9 57.1 64.7 69.4

Smooth

48.7870 49.6755 51.1521 52.1675 53.6181 57.6509 66.6199 72.4105 72.5679 62.7837 57.9154 55.5108 54.4920 53.2117 52.8118 52.0929 52.5894 55.7388 63.7189 68.7087 67.9722 59.4594 55.4037 53.4103 52.6818 51.8659 51.8002 51.6271 52.1643 55.0424 63.6079 68.6702 67.9561 59.1021 55.6210 53.7881 53.0479 52.4502 52.7444 52.8747 53.7689 57.2790 65.4702 70.6713

Predict

49.3853 50.4303 52.0132 53.0746 54.5132 58.5541 67.5607 73.3280 73.3777 63.2634 58.1732 55.6485 54.5392 53.1621 52.6957 51.9302 52.4439 55.6209 63.5782 68.5838 67.8890 59.4361 55.3468 53.3536 52.6356 51.8071 51.7290 51.5549 52.0890 54.9676 63.6199 68.6823 67.9727 59.1487 55.7003 53.8609 53.1211 52.5294 52.8467 53.0029 53.9188 57.4374 65.6357 70.8095

Error

4.11470 2.56966 1.18677 -0.57458 -1.11323 -2.05414 -2.26072 -2.62800 -6.47768 -5.06337 -2.87320 -2.24849 -2.43920 -1.66212 -1.19573 0.46985 0.85611 -0.12087 0.62178 1.01617 1.41104 -0.93606 -0.04680 0.24639 -0.33562 -0.30705 -0.02902 -0.05492 0.11103 2.13244 -0.01988 0.11774 0.92727 0.95133 -0.10032 0.03912 0.17892 0.57055 0.65329 0.49714 -0.01879 -0.33743 -0.93567 -1.40954

١٨٨

45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

70.3 62.6 57.9 55.8 54.8 54.2 54.6 54.3 54.8 58.1 68.1 73.3 75.5 66.4 60.5 57.7

69.8908 60.7552 57.1925 55.5181 54.8764 54.3244 54.5372 54.5298 55.1925 58.6054 66.7739 72.4056 72.3385 63.6729 60.0395 58.3023

69.9719 60.8370 57.3348 55.6775 55.0383 54.4749 54.6769 54.6661 55.3155 58.7141 66.8698 72.5622 72.5236 63.9378 60.3781 58.6338

0.32815 1.76302 0.56523 0.12253 -0.23826 -0.27486 -0.07694 -0.36612 -0.51551 -0.61410 1.23016 0.73784 2.97638 2.46217 0.12191 -0.93381

Row

Period

Forecast

Lower

Upper

1 2 3 4 5 6 7 8 9 10 11 12

61 62 63 64 65 66 67 68 69 70 71 72

57.8102 57.3892 57.8332 57.9307 58.8311 62.7415 72.1849 78.1507 78.5092 68.6689 63.9258 61.8189

55.0645 54.5864 54.9687 55.0005 55.8313 59.6686 69.0354 74.9215 75.1976 65.2721 60.4414 58.2446

60.5558 60.1921 60.6977 60.8609 61.8309 65.8145 75.3344 81.3798 81.8208 72.0657 67.4103 65.3933

Winters' Trend and Seasonal Smoothing

Actual

80

Predicted Forecast Actual Predicted Forecast

Food

70

60

Smoothing Constants Alpha (level): 0.200 Gamma (trend):0.200 Delta (season):0.200

50

MAPE: MAD: MSD:

0

10

20

30

40

50

60

1.88377 1.12068 2.86696

70

Time

:‫ﻣﻼﺣﻈﺎت‬ ‫ ﺛﺎﺑﺖ‬γ ‫ ﺛﺎﺑﺖ اﻟﺘﻤﻬﻴﺪ اﻟﻜﻠﻲ و‬α ‫ﻟﺘﻄﺒﻴﻖ ﻃﺮﻳﻘﺔ وﻧﺘﺮز ﻧﺤﺘﺎج اﻟﻲ إﺧﺘﻴﺎر ﻗﻴﻢ ﺛﻼﺛﺔ ﻣﻌﺎﻟﻢ هﻲ‬ ‫ ﻓﻲ ﺛﻼﺛﺔ‬Optimization ‫ ﺛﺎﺑﺖ ﺗﻤﻬﻴﺪ اﻟﻤﻮﺳﻤﻴﺔ وهﺬﻩ ﻋﻤﻠﻴﺔ أﻓﻀﻠﻴﺔ‬β ‫ﺗﻤﻬﻴﺪ اﻹﻧﺠﺮاف و‬ ١٨٩

‫اﺑﻌﺎد ) ﻓﻀﺎء اﻟﻤﻌﺎﻟﻢ ( إﻣﺎ أن ﻧﺘﺮك ﻟﻠﺒﺮﻧﺎﻣﺞ اﻹﺣﺼﺎﺋﻲ اﻟﻤﺴﺘﺨﺪم ﺣﺴﺎﺑﻬﺎ ﺗﻠﻘﺎﺋﻴﺎ ﺑﺈﺳﺘﺨﺪام‬ ‫ﺧﻮارزﻣﺎت ﻏﻴﺮ ﺧﻄﻴﺔ ﻣﺒﻨﻴﺔ داﺧﻞ اﻟﺒﺮﻧﺎﻣﺞ أو ﻧﻘﻮم ﻧﺤﻦ ﺑﺈﻣﺪاد اﻟﺒﺮﻧﺎﻣﺞ ﺑﺘﻠﻚ اﻟﻘﻴﻢ‪.‬‬ ‫ﺗﻤﺮﻳﻦ‪ :‬ﻓﻲ اﻷﻣﺜﻠﺔ اﻟﺴﺎﺑﻘﺔ اﺧﺬﻧﺎ ‪ . α = γ = β = 0.2‬ﺛﺒﺖ ﻓﻲ آﻞ ﻣﺮة ﻣﻌﻠﻤﻴﻦ وﻏﻴﺮ اﻟﺜﺎﻟﺚ‬ ‫ﺣﺘﻲ ﺗﺤﺼﻞ ﻋﻠﻲ أﻗﻞ ‪MSD‬‬ ‫ﻣﺜﺎل ﻋﻤﻠﻲ ﻟﺒﻨﺎء ﻧﻤﻮذج ﺗﻨﺒﺆ‪:‬‬ ‫ﺳﻮف ﻧﺴﺘﺨﺪم ورﻗﺔ اﻟﻌﻤﻞ ‪ CPI.MTW‬ﻣﻦ ﻣﺠﻤﻮﻋﺔ اﻟﺒﻴﺎﻧﺎت ﻟﻠﺒﺮﻧﺎﻣﺞ ‪MINITAB‬‬ ‫‪MTB > Retrieve‬‬

‫‪'C:\MTBWIN\STUDENT9\CPI.MTW'.‬‬

‫ﺳﻮف ﻧﺄﺧﺬ اﻟﻤﺘﻐﻴﺮ ‪CPIChange‬‬ ‫‪2.9‬‬ ‫‪6.2‬‬ ‫‪13.5‬‬ ‫‪3.6‬‬

‫‪1.6‬‬ ‫‪3.2‬‬ ‫‪11.3‬‬ ‫‪1.9‬‬

‫‪1.3‬‬ ‫‪5.7‬‬ ‫‪6.5‬‬ ‫‪4.3‬‬ ‫‪4.2‬‬

‫‪1.3‬‬ ‫‪4.4‬‬ ‫‪7.6‬‬ ‫‪3.6‬‬ ‫‪3.0‬‬

‫‪1.0‬‬ ‫‪4.2‬‬ ‫‪9.1‬‬ ‫‪6.2‬‬ ‫‪4.8‬‬

‫‪1.0‬‬ ‫‪5.5‬‬ ‫‪5.8‬‬ ‫‪3.2‬‬ ‫‪5.4‬‬

‫‪CPIChnge‬‬ ‫‪1.7‬‬ ‫‪3.1‬‬ ‫‪11.0‬‬ ‫‪10.3‬‬ ‫‪4.1‬‬

‫ﻧﺮﺳﻢ اﻟﻤﺘﺴﻠﺴﻠﺔ وﻧﻮﺟﺪ اﻟﺘﺮﺑﻄﺎت اﻟﺬاﺗﻴﺔ واﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﺠﺰﺋﻴﺔ اﻟﻌﻴﻨﻴﺔ‬ ‫‪MTB > TSPlot‬‬

‫‪14‬‬ ‫‪12‬‬ ‫‪10‬‬

‫‪6‬‬

‫‪CPIChnge‬‬

‫‪8‬‬

‫‪4‬‬ ‫‪2‬‬ ‫‪0‬‬ ‫‪30‬‬

‫‪20‬‬

‫‪25‬‬

‫‪10‬‬

‫‪15‬‬

‫‪Index‬‬

‫‪5‬‬

‫‪MTB > %acf c2‬‬ ‫‪Autocorrelation Function for CPIChnge‬‬

‫‪8‬‬

‫‪7‬‬

‫‪5‬‬

‫‪6‬‬

‫‪4‬‬

‫‪Corr‬‬

‫‪Lag‬‬

‫‪LBQ‬‬

‫‪T‬‬

‫‪41.23‬‬

‫‪8 -0.16 -0.51‬‬

‫‪١٩٠‬‬

‫‪3‬‬

‫‪2‬‬

‫‪Corr‬‬

‫‪1‬‬

‫‪Lag‬‬

‫‪LBQ‬‬

‫‪T‬‬

‫‪22.42‬‬ ‫‪30.32‬‬ ‫‪33.59‬‬ ‫‪36.24‬‬ ‫‪38.97‬‬ ‫‪40.05‬‬ ‫‪40.07‬‬

‫‪1 0.79 4.53‬‬ ‫‪2 0.46 1.77‬‬ ‫‪3 0.29 1.02‬‬ ‫‪4 0.26 0.88‬‬ ‫‪5 0.26 0.86‬‬ ‫‪6 0.16 0.52‬‬ ‫‪7 -0.02 -0.06‬‬

‫‪Autocorrelation‬‬

‫‪1.0‬‬ ‫‪0.8‬‬ ‫‪0.6‬‬ ‫‪0.4‬‬ ‫‪0.2‬‬ ‫‪0.0‬‬ ‫‪-0.2‬‬ ‫‪-0.4‬‬ ‫‪-0.6‬‬ ‫‪-0.8‬‬ ‫‪-1.0‬‬

MTB > %pacf c2 Partial Autocorrelation

Partial Autocorrelation Function for CPIChnge 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

1

2

3

4

Lag PAC 1 2 3 4 5 6 7

0.79 -0.42 0.35 -0.04 0.08 -0.28 -0.05

5

6

T

Lag PAC

T

4.53 -2.44 2.01 -0.24 0.43 -1.62 -0.29

8 -0.09

-0.52

7

8

ARMA(١،١) ‫ﻣﻦ اﻧﻤﺎط اﻟﺘﺮﺑﻄﺎت اﻟﺬاﺗﻴﺔ واﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﺠﺰﺋﻴﺔ اﻟﻌﻴﻨﻴﺔ ﻗﺪ ﻳﻜﻮن اﻟﻨﻤﻮذج‬ ‫ ﺗﻨﺒﺆات ﻟﻠﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ‬٥ ‫ اﻷواﻣﺮ اﻟﺘﺎﻟﻴﺔ ﺗﻄﺒﻖ اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح وﺗﻮﻟﺪ‬،‫ﻳﻨﻄﺒﻖ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ‬ MTB > SUBC> SUBC> SUBC> SUBC> SUBC> SUBC>

arima 1 0 1 c2; fore 5 c3 c4 c5; gser; gacf; gpacf; ghist; gnormal.

ARIMA Model ARIMA model for CPIChnge Estimates at each iteration Iteration SSE Parameters 0 323.251 0.100 0.100 4.522 1 200.616 0.250 -0.050 3.745 2 182.146 0.184 -0.200 4.067 3 163.067 0.135 -0.350 4.308 4 142.864 0.107 -0.500 4.434 5 121.402 0.111 -0.650 4.407 6 99.668 0.150 -0.800 4.197 7 77.036 0.268 -0.950 3.590 8 67.550 0.418 -0.956 2.828 9 62.802 0.568 -0.964 2.062 10 62.108 0.637 -0.973 1.687 11 62.030 0.644 -0.979 1.619 12 62.003 0.647 -0.982 1.584 13 61.996 0.651 -0.985 1.549 14 61.996 0.651 -0.986 1.539 Unable to reduce sum of squares any further Final Estimates of Parameters

١٩١

Type AR 1 MA 1 Constant Mean

Coef 0.6513 -0.9857 1.5385 4.412

StDev 0.1434 0.0516 0.4894 1.403

T 4.54 -19.11 3.14

Number of observations: 33 Residuals: SS = 61.8375 (backforecasts excluded) MS = 2.0613 DF = 30 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 9.6(DF=10) 17.0(DF=22) * (DF= *)

48 * (DF= *)

Forecasts from period 33 Period 34 35 36 37 38

95 Percent Limits Lower Upper 0.3216 5.9507 -1.8180 8.9801 -2.3061 10.0476 -2.4192 10.5380 -2.4201 10.7848

Forecast 3.1362 3.5810 3.8708 4.0594 4.1823

zt = 1.54 + 0.65zt −1 + at − 0.99at −1 , at

Actual

‫أي ان اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح هﻮ‬

N ( 0, 2.06 )

‫ هﻲ‬t ‫ﻣﻘﺪرات اﻟﻤﻌﺎﻟﻢ وإﻧﺤﺮاﻓﺎﺗﻬﺎ اﻟﻤﻌﻴﺎرﻳﺔ وﻗﻴﻤﺔ إﺧﺘﺒﺎر‬ ˆ ˆ φ1 = 0.6513, s.e. φ1 = 0.1434, t = 4.54

( ) θˆ = −0.9857, s.e. (θˆ ) = 0.0516, t = −19.11 δˆ = 1.5385, s.e. (δˆ ) = 0.4894, t = 3.14 1

1

σˆ 2 = 2.0613, with d . f . = 30

.‫ﻧﻼﺣﻆ ان ﺟﻤﻴﻊ اﻟﻤﻌﺎﻟﻢ ﻣﻌﻨﻮﻳﺔ‬

:‫اﻵن ﻧﻔﺤﺺ اﻟﺒﻮاﻗﻲ‬

ACF of Residuals for CPIChnge (with 95% confidence limits for the autocorrelations) 1.0 0.8

Autocorrelation

0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

١٩٢

PACF of Residuals for CPIChnge (with 95% confidence limits for the partial autocorrelations) 1.0

Partial Autocorrelation

0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0 1

2

3

4

5

6

7

8

Lag

‫أﻧﻤﺎط اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﺒﻮاﻗﻲ ﺗﺪل ﻋﻠﻰ أن اﻟﺒﻮاﻗﻲ ﺗﺘﺒﻊ ﺗﻮزﻳﻊ ﺿﺠﺔ‬ :‫ ﻟﻨﻔﺤﺺ ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬،‫ﺑﻴﻀﺎء أي ﻏﻴﺮ ﻣﺘﺮاﺑﻄﺔ‬ ‫رﺳﻢ اﻟﻤﺪرج اﻟﺘﻜﺮاري‬

Histogram of the Residuals (response is CPIChnge) 8 7

Frequency

6 5 4 3 2

١٩٣

.‫ﻳﺒﺪو ﻣﺘﻨﺎﻇﺮ ﺑﻌﺾ اﻟﺸﻴﺊ‬ :‫ﻟﻨﻨﻈﺮ إﻟﻰ ﻣﺨﻄﻂ اﻹﺣﺘﻤﺎل اﻟﻄﺒﻴﻌﻲ ﻟﻠﺒﻮاﻗﻲ‬

Normal Probability Plot of the Residuals (response is CPIChnge) 4 3

Residual

2 1 0 -1 -2 -3 -2

-1

0

1

2

Normal Score

.‫ﻧﺴﺘﻄﻴﻊ أن ﻧﻘﻮل ان اﻟﺒﻮاﻗﻲ ﻃﺒﻴﻌﻴﺔ ﺗﻘﺮﻳﺒﺎ‬ .‫ ﺗﻨﺒﺆات ﻟﻠﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ‬٥ ‫اﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻟﻠﻤﺘﺴﻠﺴﺔ ﻣﻊ‬

Time Series Plot for CPIChnge (with forecasts and their 95% confidence limits)

CPIChnge

10

5

0

١٩٤

:‫ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ آﺎﻟﺘﺎﻟﻲ‬AR(٢) ‫دﻋﻨﺎ ﻧﺤﺎول ﺗﻄﺒﻴﻖ ﻧﻤﻮذج‬ MTB > arima 2 0 0 c2 Type AR 1 AR 2 Constant Mean

Coef 1.1872 -0.4657 1.3270 4.765

StDev 0.1625 0.1624 0.2996 1.076

T 7.31 -2.87 4.43

Number of observations: 33 Residuals: SS = 88.6206 (backforecasts excluded) MS = 2.9540 DF = 30 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 19.8(DF=10) 25.4(DF=22) * (DF= *)

zt = 1.33 + 1.187 zt −1 − 0.4657 zt −1 + at , at

N ( 0, 2.95)

48 * (DF= *)

‫أي ان اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح هﻮ‬

‫ هﻲ‬t ‫ﻣﻘﺪرات اﻟﻤﻌﺎﻟﻢ وإﻧﺤﺮاﻓﺎﺗﻬﺎ اﻟﻤﻌﻴﺎرﻳﺔ وﻗﻴﻤﺔ إﺧﺘﺒﺎر‬

( ) φˆ = −0.4657, s.e. (θˆ ) = 0.1624, t = −2.87 δˆ = 1.327, s.e. (δˆ ) = 0.2996, t = 4.43

φˆ1 = 1.1872, s.e. φˆ1 = 0.1625, t = 7.31 2

1

σˆ 2 = 2.954, with d . f . = 30

‫ﻟﻨﻨﻈﺮ إﻟﻰ اﻹﺧﺘﺒﺎر‬

H 0 : φ2 = 0 H 1 : φ2 ± 0

‫اﻹﺣﺼﺎﺋﺔ‬ t0 =

φˆ

2

( )

s.e. φˆ2

=

−0.4657 = −2.8676 0.1624

‫ ﻟﻬﺎ ﺑﺎﻷﻣﺮ‬P-value ‫ﻧﻮﺟﺪ اﻟـ‬ MTB > cdf -2.8676; SUBC> t 30.

Cumulative Distribution Function

١٩٥

‫‪Student's t distribution with 30 DF‬‬ ‫)‪P( X <= x‬‬ ‫‪0.0037‬‬

‫‪x‬‬ ‫‪-2.8676‬‬

‫أي اﻟـ ‪ P-value‬ﻟﻬﺎ ﺗﺴﺎوي ‪ ٠٫٠٠٣٧‬وهﻲ أﻗﻞ ﻣﻦ ‪ ٠٫٠٥‬أي ﻻﻧﺮﻓﺾ ان ‪ φ2 = 0‬وﺑﺎﻟﺘﺎﻟﻲ‬ ‫ﻧﺮﻓﺾ اﻟﻨﻤﻮذج )‪. AR(٢‬‬ ‫ﺗﻤﺮﻳﻦ‪:‬‬ ‫ﺣﺎول ﺗﻄﺒﻴﻖ ﻧﻤﺎذج اﺧﺮى ﻣﻨﺎﺳﺒﺔ ﻋﻠﻰ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﺴﺎﺑﻘﺔ وإﺧﺘﺎر أﻓﻀﻞ ﻧﻤﻮذج‪ ،‬أﺟﺮي‬ ‫اﻹﺧﺘﺒﺎرات اﻟﻤﻨﺎﺳﺒﺔ واﺳﺘﺨﺪم اﻳﻀﺎ اﻟﻤﻌﻴﺎر ‪. AIC‬‬ ‫ﻣﺜﺎل ﻋﻤﻠﻲ ﺁﺧﺮ ﻟﺒﻨﺎء ﻧﻤﻮذج ﺗﻨﺒﺆ‪:‬‬ ‫ﺳﻮف ﻧﺤﺎول ﺑﻨﺎء ﻧﻤﻮذج ﻟﻠﻤﺘﺴﻠﺴﻠﺔ‬ ‫)‪z(t‬‬ ‫‪-103.2‬‬ ‫‪-391.4‬‬ ‫‪-836.1‬‬ ‫‪-1307.3‬‬ ‫‪-1736.9‬‬ ‫‪-2097.5‬‬ ‫‪-2363.4‬‬ ‫‪-2721.4‬‬ ‫‪-3353.3‬‬ ‫‪-4153.0‬‬ ‫‪-5068.7‬‬ ‫‪-5848.8‬‬ ‫‪-6387.3‬‬ ‫‪-6834.5‬‬ ‫‪-7259.9‬‬ ‫‪-7891.4‬‬ ‫‪-8791.2‬‬ ‫‪-9878.4‬‬ ‫‪-11071.9‬‬ ‫‪-12157.5‬‬ ‫‪-13190.3‬‬ ‫‪-14196.3‬‬ ‫‪-15335.9‬‬ ‫‪-16697.1‬‬ ‫‪-18155.6‬‬ ‫‪-19547.0‬‬ ‫‪-20827.7‬‬ ‫‪-21908.3‬‬

‫‪-76.7‬‬ ‫‪-339.9‬‬ ‫‪-766.7‬‬ ‫‪-1242.9‬‬ ‫‪-1679.0‬‬ ‫‪-2055.4‬‬ ‫‪-2328.5‬‬ ‫‪-2651.9‬‬ ‫‪-3253.4‬‬ ‫‪-4028.2‬‬ ‫‪-4937.2‬‬ ‫‪-5760.6‬‬ ‫‪-6317.8‬‬ ‫‪-6773.3‬‬ ‫‪-7193.3‬‬ ‫‪-7784.7‬‬ ‫‪-8649.8‬‬ ‫‪-9713.3‬‬ ‫‪-10903.2‬‬ ‫‪-12011.7‬‬ ‫‪-13039.8‬‬ ‫‪-14051.7‬‬ ‫‪-15158.3‬‬ ‫‪-16492.7‬‬ ‫‪-17951.5‬‬ ‫‪-19356.3‬‬ ‫‪-20655.7‬‬ ‫‪-21762.1‬‬

‫‪-52.8‬‬ ‫‪-291.3‬‬ ‫‪-698.2‬‬ ‫‪-1177.4‬‬ ‫‪-1620.8‬‬ ‫‪-2010.0‬‬ ‫‪-2290.6‬‬ ‫‪-2589.5‬‬ ‫‪-3156.1‬‬ ‫‪-3906.6‬‬ ‫‪-4805.0‬‬ ‫‪-5663.8‬‬ ‫‪-6244.3‬‬ ‫‪-6711.8‬‬ ‫‪-7131.7‬‬ ‫‪-7683.2‬‬ ‫‪-8512.3‬‬ ‫‪-9552.0‬‬ ‫‪-10734.0‬‬ ‫‪-11863.9‬‬ ‫‪-12889.7‬‬ ‫‪-13910.1‬‬ ‫‪-14986.4‬‬ ‫‪-16290.9‬‬ ‫‪-17745.2‬‬ ‫‪-19161.7‬‬ ‫‪-20478.0‬‬ ‫‪-21614.0‬‬

‫‪-33.1‬‬ ‫‪-246.0‬‬ ‫‪-631.6‬‬ ‫‪-1111.3‬‬ ‫‪-1561.5‬‬ ‫‪-1960.6‬‬ ‫‪-2250.5‬‬ ‫‪-2533.0‬‬ ‫‪-3060.8‬‬ ‫‪-3788.7‬‬ ‫‪-4673.3‬‬ ‫‪-5558.0‬‬ ‫‪-6167.9‬‬ ‫‪-6649.0‬‬ ‫‪-7073.1‬‬ ‫‪-7586.4‬‬ ‫‪-8379.1‬‬ ‫‪-9394.6‬‬ ‫‪-10564.0‬‬ ‫‪-11713.8‬‬ ‫‪-12740.8‬‬ ‫‪-13769.9‬‬ ‫‪-14819.8‬‬ ‫‪-16091.5‬‬ ‫‪-17535.7‬‬ ‫‪-18965.1‬‬ ‫‪-20294.5‬‬ ‫‪-21463.7‬‬ ‫‪-22474.4‬‬

‫وﻟﻬﺎ اﻟﺮﺳﻢ اﻟﺰﻣﻨﻲ اﻟﺘﺎﻟﻲ‪:‬‬

‫‪١٩٦‬‬

‫‪-19.2‬‬ ‫‪-204.0‬‬ ‫‪-566.7‬‬ ‫‪-1044.2‬‬ ‫‪-1500.0‬‬ ‫‪-1906.9‬‬ ‫‪-2210.8‬‬ ‫‪-2482.0‬‬ ‫‪-2968.6‬‬ ‫‪-3675.0‬‬ ‫‪-4542.3‬‬ ‫‪-5444.4‬‬ ‫‪-6090.4‬‬ ‫‪-6584.6‬‬ ‫‪-7015.6‬‬ ‫‪-7495.3‬‬ ‫‪-8249.2‬‬ ‫‪-9239.6‬‬ ‫‪-10392.1‬‬ ‫‪-11560.2‬‬ ‫‪-12593.4‬‬ ‫‪-13629.3‬‬ ‫‪-14657.5‬‬ ‫‪-15895.8‬‬ ‫‪-17324.4‬‬ ‫‪-18766.2‬‬ ‫‪-20107.3‬‬ ‫‪-21310.9‬‬ ‫‪-22335.1‬‬

‫‪-9.1‬‬ ‫‪-165.9‬‬ ‫‪-504.8‬‬ ‫‪-975.1‬‬ ‫‪-1436.5‬‬ ‫‪-1851.1‬‬ ‫‪-2173.3‬‬ ‫‪-2437.6‬‬ ‫‪-2880.4‬‬ ‫‪-3564.3‬‬ ‫‪-4412.0‬‬ ‫‪-5323.6‬‬ ‫‪-6011.8‬‬ ‫‪-6518.9‬‬ ‫‪-6957.1‬‬ ‫‪-7411.2‬‬ ‫‪-8124.0‬‬ ‫‪-9086.3‬‬ ‫‪-10219.0‬‬ ‫‪-11402.0‬‬ ‫‪-12447.5‬‬ ‫‪-13486.3‬‬ ‫‪-14499.7‬‬ ‫‪-15705.0‬‬ ‫‪-17113.5‬‬ ‫‪-18564.2‬‬ ‫‪-19919.5‬‬ ‫‪-21154.4‬‬ ‫‪-22195.6‬‬

‫‪-2.5‬‬ ‫‪-132.4‬‬ ‫‪-446.4‬‬ ‫‪-905.3‬‬ ‫‪-1371.8‬‬ ‫‪-1794.5‬‬ ‫‪-2136.4‬‬ ‫‪-2398.7‬‬ ‫‪-2797.6‬‬ ‫‪-3456.9‬‬ ‫‪-4281.7‬‬ ‫‪-5197.8‬‬ ‫‪-5931.6‬‬ ‫‪-6453.5‬‬ ‫‪-6896.0‬‬ ‫‪-7332.9‬‬ ‫‪-8004.6‬‬ ‫‪-8936.4‬‬ ‫‪-10047.2‬‬ ‫‪-11238.9‬‬ ‫‪-12302.5‬‬ ‫‪-13339.6‬‬ ‫‪-14345.9‬‬ ‫‪-15518.4‬‬ ‫‪-16904.2‬‬ ‫‪-18360.1‬‬ ‫‪-19733.5‬‬ ‫‪-20993.7‬‬ ‫‪-22053.2‬‬

O r ig in a l T im e S e r ie s

-1 0 0 0 0

-2 0 0 0 0

In d e x

50

100

150

200

:‫اﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ واﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﺠﺰﺋﻴﺔ هﻲ‬

Autocorrelation Function for z(t) Autocorrelation

1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

10 Lag Corr 1 2 3 4 5 6 7 8 9 10 11 12

T

LBQ

0.9813.93 196.91 0.97 8.00 388.73 0.95 6.15 575.44 0.94 5.15 757.06 0.92 4.50 933.59 0.91 4.041105.06 0.89 3.681271.47 0.88 3.391432.85 0.86 3.151589.23 0.84 2.951740.66 0.83 2.781887.17 0.81 2.622028.81

Lag Corr 13 14 15 16 17 18 19 20 21 22 23 24

20 T

LBQ

2.49 2165.64 2.37 2297.73 2.25 2425.14 2.15 2547.94 2.06 2666.22 1.97 2780.05 1.89 2889.51 1.82 2994.69 1.75 3095.67 1.68 3192.54 1.62 3285.39 1.56 3374.32

0.80 0.78 0.76 0.75 0.73 0.72 0.70 0.68 0.67 0.65 0.64 0.62

30

Lag Corr 25 26 27 28 29 30 31 32 33 34 35 36

0.61 0.59 0.58 0.56 0.55 0.53 0.52 0.50 0.49 0.47 0.46 0.45

T

LBQ

1.50 3459.42 1.45 3540.78 1.40 3618.50 1.35 3692.67 1.30 3763.38 1.26 3830.74 1.21 3894.84 1.17 3955.76 1.13 4013.61 1.09 4068.47 1.05 4120.44 1.01 4169.59

40

Lag Corr 37 38 39 40 41 42 43 44 45 46 47 48

0.43 0.42 0.41 0.39 0.38 0.36 0.35 0.34 0.32 0.31 0.30 0.28

T

LBQ

50

Lag Corr

T

LBQ

49 0.27 0.594589.24 50 0.26 0.564607.24

0.98 4216.02 0.94 4259.81 0.91 4301.04 0.87 4339.79 0.84 4376.14 0.81 4410.16 0.77 4441.93 0.74 4471.52 0.71 4498.99 0.68 4524.43 0.65 4547.90 0.62 4569.48

P artial A utocorrelation Function for z(t) Partial Autocorrelation

z(t)

0

1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

10

20

30

40

Lag

PAC

T

Lag

PAC

T

Lag

PAC

T

Lag

PAC

T

1 2 3 4 5 6 7 8 9 10 11 12

0.98 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01

13.93 -0.17 -0.17 -0.16 -0.16 -0.16 -0.15 -0.15 -0.15 -0.14 -0.14 -0.13

13 14 15 16 17 18 19 20 21 22 23 24

-0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01

-0.13 -0.12 -0.12 -0.11 -0.11 -0.11 -0.11 -0.10 -0.10 -0.09 -0.09 -0.09

25 26 27 28 29 30 31 32 33 34 35 36

-0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01

-0.09 -0.09 -0.09 -0.08 -0.08 -0.08 -0.08 -0.08 -0.08 -0.08 -0.09 -0.09

37 38 39 40 41 42 43 44 45 46 47 48

-0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01 -0.01

-0.09 -0.09 -0.10 -0.10 -0.11 -0.11 -0.12 -0.12 -0.12 -0.13 -0.13 -0.13

50 Lag

PAC

T

49 -0.01 50 -0.01

-0.14 -0.14

.‫ ﻏﻴﺮ ﻣﺴﺘﻘﺮة ﻓﻲ اﻟﻤﺘﻮﺳﻂ‬zt ‫واﺿﺢ ﺟﺪا ان اﻟﻤﺘﺴﻠﺴﻠﺔ‬ ‫ وﻧﺮﺳﻤﻬﺎ‬wt = zt − zt −1 ‫ﻧﺄﺧﺬ اﻟﻔﺮوق اﻻوﻟﻰ‬ ١٩٧

F irs t D if f e re n c e s w (t)= z (t)-z (t-1 )

w(t)

0

-1 0 0

-2 0 0 In d e x

50

100

150

200

:‫اﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ واﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﺠﺰﺋﻴﺔ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﻔﺮﻗﺔ هﻲ‬ Autocorrelation

Autocorrelation Function for w(t) 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

5 Lag Corr 1 2 3 4 5 6 7 8 9 10 11 12

T

LBQ

0.9913.93 197.10 0.97 7.98 389.03 0.95 6.12 574.74 0.93 5.10 753.55 0.91 4.43 924.81 0.89 3.941087.93 0.86 3.561242.44 0.83 3.251388.10 0.81 2.981524.83 0.78 2.761652.58 0.75 2.561771.42 0.72 2.381881.49

15

Lag Corr 13 14 15 16 17 18 19 20 21 22 23 24

0.69 0.66 0.63 0.60 0.57 0.54 0.52 0.49 0.47 0.45 0.43 0.41

T

LBQ

2.21 1983.03 2.07 2076.38 1.93 2161.98 1.81 2240.33 1.70 2311.99 1.60 2377.42 1.50 2437.13 1.41 2491.57 1.33 2541.23 1.26 2586.61 1.19 2628.22 1.13 2666.48

25 Lag Corr 25 26 27 28 29 30 31 32 33 34 35 36

0.39 0.38 0.37 0.36 0.35 0.34 0.34 0.34 0.34 0.34 0.34 0.34

T

LBQ

1.08 2701.91 1.03 2734.99 1.00 2766.16 0.96 2795.84 0.94 2824.46 0.92 2852.35 0.91 2879.79 0.90 2907.03 0.89 2934.32 0.89 2961.85 0.89 2989.82 0.89 3018.34

35 Lag Corr 37 38 39 40 41 42 43 44 45 46 47 48

0.34 0.35 0.35 0.35 0.35 0.35 0.35 0.34 0.34 0.33 0.33 0.32

T

45 LBQ

0.90 3047.50 0.90 3077.32 0.90 3107.68 0.90 3138.42 0.90 3169.33 0.89 3200.21 0.88 3230.81 0.87 3260.99 0.85 3290.64 0.84 3319.59 0.82 3347.55 0.79 3374.17

Lag Corr

T

LBQ

49 0.31 0.763399.16

Partial Autocorrelation

P a rtia l A u to c o rre la tio n F u n c tio n fo r w (t) 1 .0 0 .8 0 .6 0 .4 0 .2 0 .0 -0 .2 -0 .4 -0 .6 -0 .8 -1 .0

5 Lag

P AC

T

1 2 3 4 5 6 7 8 9 10 11 12

0 .9 9 -0 .1 4 -0 .1 1 -0 .0 7 -0 .0 6 -0 .0 7 -0 .0 6 -0 .0 3 -0 .0 3 -0 .0 3 -0 .0 2 -0 .0 2

1 3 .9 3 -1 .9 7 -1 .5 1 -0 .9 6 -0 .9 0 -0 .9 7 -0 .7 8 -0 .4 8 -0 .3 7 -0 .3 7 -0 .3 4 -0 .2 2

15

Lag

P AC

T

1 3 -0 .0 1 1 4 -0 .0 0 1 5 0 .0 1 1 6 0 .0 2 1 7 0 .0 2 1 8 0 .0 1 1 9 0 .0 1 2 0 0 .0 1 2 1 0 .0 1 2 2 0 .0 3 2 3 0 .0 2 2 4 0 .0 2

-0 .1 6 -0 .0 2 0 .1 0 0 .2 3 0 .2 5 0 .1 0 0 .1 2 0 .1 1 0 .1 6 0 .3 5 0 .3 1 0 .2 7

25 Lag

35

45

P AC

T

Lag

P AC

T

2 5 0 .0 4 2 6 0 .0 4 2 7 0 .0 3 2 8 0 .0 4 2 9 0 .0 4 3 0 0 .0 2 3 1 0 .0 2 3 2 0 .0 1 3 3 0 .0 1 3 4 0 .0 1 3 5 0 .0 0 3 6 -0 .0 0

0 .5 2 0 .6 2 0 .4 7 0 .5 5 0 .5 6 0 .3 4 0 .2 2 0 .1 6 0 .1 8 0 .1 5 0 .0 4 -0 .0 7

37 38 39 40 41 42 43 44 45 46 47 48

-0 .0 1 -0 .0 2 -0 .0 4 -0 .0 4 -0 .0 4 -0 .0 4 -0 .0 3 -0 .0 0 0 .0 1 -0 .0 1 -0 .0 4 -0 .0 5

-0 .0 7 -0 .2 9 -0 .6 3 -0 .6 0 -0 .5 4 -0 .5 5 -0 .4 2 -0 .0 7 0 .0 8 -0 .1 8 -0 .5 6 -0 .7 6

Lag

P AC

T

4 9 -0 .0 3

-0 .4 8

.‫ ﻻﺗﺰال ﻏﻴﺮ ﻣﺴﺘﻘﺮة ﻓﻲ اﻟﻤﺘﻮﺳﻂ‬wt ‫واﺿﺢ ﺟﺪا ان اﻟﻤﺘﺴﻠﺴﻠﺔ‬ ‫ )ﻻﺣﻆ ان هﺬا اﻟﻔﺮق اﻟﺜﺎﻧﻲ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ اﻷﺻﻠﻴﺔ( وﻧﺮﺳﻤﻬﺎ‬yt = wt − wt −1 ‫ﻧﺄﺧﺬ اﻟﻔﺮوق اﻻوﻟﻰ‬

١٩٨

F ir s t D if f e r e n c e s

y (t)= w (t)-w (t-1 )

1 0

y(t)

5

0

-5

In d e x

5 0

1 0 0

1 5 0

2 0 0

:‫اﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ واﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﺠﺰﺋﻴﺔ ﻟﻬﺬﻩ اﻟﻤﺘﺴﻠﺴﻠﺔ اﻟﻤﻔﺮﻗﺔ هﻲ‬ Autocorrelation

A utocorrelation Function for y(t) 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

5 Lag Corr 1 2 3 4 5 6 7 8 9 10 11 12

T

LBQ

0.9313.08 173.62 0.80 6.83 303.36 0.70 4.89 401.81 0.62 3.89 479.62 0.56 3.26 542.93 0.50 2.77 593.77 0.43 2.32 632.35 0.36 1.90 659.84 0.29 1.50 677.66 0.22 1.11 687.71 0.15 0.75 692.41 0.08 0.41 693.82

15

Lag Corr

T

LBQ

25 Lag Corr

13 0.01 0.06 693.85 14 -0.05 -0.23 694.31 15 -0.09 -0.46 696.12 16 -0.14 -0.70 700.34 17 -0.19 -0.97 708.51 18 -0.25 -1.24 722.16 19 -0.31 -1.53 743.30 20 -0.37 -1.83 774.40 21 -0.44 -2.13 818.29 22 -0.51 -2.40 876.73 23 -0.56 -2.57 948.00 1030.79 24 -0.60 -2.67

T

LBQ

1124.21 25 -0.64 -2.73 1224.05 26 -0.66 -2.72 1324.28 27 -0.66 -2.62 1418.06 28 -0.63 -2.44 1500.63 29 -0.59 -2.22 1572.45 30 -0.55 -2.01 1632.42 31 -0.50 -1.80 1677.44 32 -0.43 -1.53 1708.46 33 -0.36 -1.25 1728.63 34 -0.29 -1.00 1740.51 35 -0.22 -0.76 1746.42 36 -0.16 -0.53

35 Lag Corr

T

45 LBQ

1748.28 37 -0.09 -0.30 1748.29 38 -0.01 -0.03 1749.70 39 0.08 0.26 1755.14 40 0.15 0.50 1765.20 41 0.20 0.68 1778.55 42 0.23 0.78 1793.44 43 0.24 0.82 1810.08 44 0.25 0.86 1831.09 45 0.28 0.96 1860.92 46 0.34 1.13 1903.36 47 0.40 1.34 1956.81 48 0.45 1.49

Lag Corr

T

LBQ

49 0.48 1.562017.77

Partial Autocorrelation

P artial A utocorrelation Function for y(t) 1.0 0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0

5 Lag PAC

T

0.93 -0.46 0.33 -0.14 0.12 -0.13 0.00 -0.06 -0.08 -0.04 -0.04 -0.09

13.08 -6.44 4.71 -1.97 1.72 -1.78 0.01 -0.81 -1.11 -0.55 -0.58 -1.21

1 2 3 4 5 6 7 8 9 10 11 12

15

Lag PAC 13 14 15 16 17 18 19 20 21 22 23 24

-0.05 0.05 -0.07 -0.09 -0.04 -0.08 -0.12 -0.11 -0.15 -0.07 -0.06 -0.15

25

T

Lag

PAC

T

-0.75 0.67 -0.95 -1.30 -0.55 -1.11 -1.71 -1.50 -2.12 -0.93 -0.87 -2.12

25 26 27 28 29 30 31 32 33 34 35 36

-0.03 -0.03 0.00 0.08 -0.01 -0.04 0.15 0.06 -0.00 0.04 0.03 0.04

-0.46 -0.44 0.04 1.13 -0.14 -0.56 2.17 0.88 -0.05 0.49 0.46 0.50

35 Lag PAC 37 38 39 40 41 42 43 44 45 46 47 48

0.08 0.11 0.04 0.04 -0.05 -0.05 -0.14 -0.01 0.07 0.06 -0.01 -0.09

45 T

1.16 1.56 0.51 0.52 -0.74 -0.66 -1.95 -0.10 0.92 0.79 -0.17 -1.24

Lag PAC 49

0.07

T 1.00

‫ﻧﻼﺣﻆ ﻣﻦ ﺷﻜﻞ اﻟﻤﺘﺴﻠﺴﻠﺔ و اﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ واﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﺠﺰﺋﻴﺔ اﻧﻬﺎ اﺻﺒﺤﺖ ﻣﺴﺘﻘﺮة‬ . d=٢ ‫ﻓﻲ اﻟﻤﺘﻮﺳﻂ اي ان‬ ‫ﻣﻦ اﻧﻤﺎط اﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ واﻟﺘﺮاﺑﻄﺎت اﻟﺬاﺗﻴﺔ اﻟﺠﺰﺋﻴﺔ ﻧﺮى اﻧﻬﺎ ﺗﺘﺨﺎﻣﺪ ﻣﻦ اﻟﺘﺨﻠﻒ اﻷول ﻣﻤﺎ‬ ‫ وﺳﻮف ﻧﻄﺒﻖ هﺬا اﻟﻨﻤﻮذج ﺑﺎﻷﻣﺮ‬zt ‫ ﻟﻠﻤﺘﺴﻠﺴﻠﺔ اﻷﺻﻠﻴﺔ‬ARIMA(١،٢،١) ‫ﻳﺮﺷﺢ ﻧﻤﻮذج‬ MTB > ARIMA 1 2 1 'z(t)' 'RESI2' 'FITS2'; SUBC> NoConstant; SUBC> Forecast 10 c4 c5 c6; SUBC> GACF; SUBC> GPACF; SUBC> GHistogram;

١٩٩

SUBC>

GNormalplot.

ARIMA Model ARIMA model for z(t) Estimates at each iteration Iteration SSE Parameters 0 2462.77 0.100 0.100 1 1345.58 0.250 -0.050 2 1170.63 0.203 -0.200 3 984.83 0.182 -0.350 4 782.47 0.200 -0.500 5 560.15 0.278 -0.650 6 363.93 0.428 -0.765 7 259.20 0.578 -0.814 8 202.76 0.728 -0.842 9 185.51 0.861 -0.859 10 185.36 0.873 -0.860 11 185.36 0.875 -0.860 12 185.36 0.875 -0.860 Relative change in each estimate less than Final Estimates of Parameters Type Coef StDev AR 1 0.8749 0.0353 MA 1 -0.8599 0.0357

0.0010

T 24.75 -24.12

Differencing: 2 regular differences Number of observations: Original series 200, after differencing 198 Residuals: SS = 183.717 (backforecasts excluded) MS = 0.937 DF = 196 Modified Box-Pierce (Ljung-Box) Chi-Square statistic Lag 12 24 36 Chi-Square 3.9(DF=10) 13.0(DF=22) 33.1(DF=34)

48 46.0(DF=46)

Forecasts from period 200 Period 201 202 203 204 205 206 207 208 209 210

Forecast -22615.2 -22757.2 -22900.4 -23044.5 -23189.5 -23335.2 -23481.5 -23628.4 -23775.8 -23923.7

95 Percent Limits Lower Upper -22617.1 -22613.3 -22764.6 -22749.9 -22917.2 -22883.5 -23075.3 -23013.7 -23238.8 -23140.1 -23407.8 -23262.6 -23582.1 -23380.9 -23761.8 -23495.0 -23946.7 -23605.0 -24136.6 -23710.7

zt = 0.875t −1 z + at − 0.859at −1 , at

N ( 0,0.937 )

Actual

‫اﻟﻨﻤﻮذج اﻟﻤﻘﺘﺮح هﻮ‬

‫ هﻲ‬t ‫وﻣﻘﺪرات اﻟﻤﻌﺎﻟﻢ وإﻧﺤﺮاﻓﺎﺗﻬﺎ اﻟﻤﻌﻴﺎرﻳﺔ وﻗﻴﻤﺔ إﺧﺘﺒﺎر‬

٢٠٠

( ) s.e. (θˆ ) = 0.0357,

φˆ1 = 0.8749, s.e. φˆ1 = 0.0353, t = 24.75 θˆ1 = −0.8599,

1

t = −24.12

σˆ 2 = 0.937, with d . f . = 196

.‫ﻧﻼﺣﻆ ان اﻟﻤﻌﺎﻟﻢ ﻣﻌﻨﻮﻳﺔ‬ :‫اﻵن ﻧﻔﺤﺺ اﻟﺒﻮاﻗﻲ‬ ACF of Residuals for z(t) (with 95% confidence limits for the autocorrelations) 1.0 0.8

Autocorrelation

0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0 5

10

15

20

25

30

35

40

45

Lag

PACF of Residuals for z(t) (with 95% confidence limits for the partial autocorrelations) 1.0

Partial Autocorrelation

0.8 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0 5

10

15

20

25

30

35

40

45

Lag

‫أﻧﻤﺎط اﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ واﻟﺘﺮاﺑﻂ اﻟﺬاﺗﻲ اﻟﺠﺰﺋﻲ ﻟﻠﺒﻮاﻗﻲ ﺗﺪل ﻋﻠﻰ أن اﻟﺒﻮاﻗﻲ ﺗﺘﺒﻊ ﺗﻮزﻳﻊ ﺿﺠﺔ‬ :‫ ﻟﻨﻔﺤﺺ ﻃﺒﻴﻌﻴﺔ اﻟﺒﻮاﻗﻲ‬،‫ﺑﻴﻀﺎء أي ﻏﻴﺮ ﻣﺘﺮاﺑﻄﺔ‬ ‫رﺳﻢ اﻟﻤﺪرج اﻟﺘﻜﺮاري‬

٢٠١

Histogram of the Residuals (response is z(t))

Frequency

20

10

0 -3

-2

-1

0

1

2

Residual

.‫ﻳﺒﺪو ﻣﺘﻨﺎﻇﺮ ﺑﻌﺾ اﻟﺸﻴﺊ‬ :‫ﻟﻨﻨﻈﺮ إﻟﻰ ﻣﺨﻄﻂ اﻹﺣﺘﻤﺎل اﻟﻄﺒﻴﻌﻲ ﻟﻠﺒﻮاﻗﻲ‬ Normal Probability Plot of the Residuals (response is z(t))

2

Residual

1

0

-1

-2

-3 -3

-2

-1

0

1

2

3

Normal Score

.‫ﻧﺴﺘﻄﻴﻊ أن ﻧﻘﻮل ان اﻟﺒﻮاﻗﻲ ﻃﺒﻴﻌﻴﺔ ﺗﻘﺮﻳﺒﺎ‬ .‫ ﻓﺘﺮات ﺗﻨﺒﺆ‬٩٥٪ ‫ ﺗﻨﺒﺆات ﻟﻠﻘﻴﻢ اﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﻣﻊ‬١٠ ‫اﻟﺮﺳﻢ اﻟﺘﺎﻟﻲ ﻟـ‬ Forecast of 20 Future values with 95% limits -22500

z(t)

-23000

-23500

-24000

0

1

2

3

4

5

6

7

8

9

Time

٢٠٢

10

٢٠٣

:‫اﻟﻤﺮاﺟﻊ‬ ١-

Abraham, B. and Ledoter, J. (1983). Statistical Methods for Forecasting, John Wiley, New York. ٢Anderson, T. W. (1971). The Statistical Analysis of Time Series, John Wiley, New York. ٣Box, G. E. P. and Jenkins, G. M. (1976). Time Series Analysis

Forecasting and Control, 2nd ed., Holden-Day, San Francisco. Brillinger, D. R. (1975). Time Series: Data Analysis and Theory, Holt, Rinehart and Winston, New York. Fuller, W. A. (1976). Introduction to Statistical Time Series, John Wiley, New York. Granger, C. W. J. and Newbold, P. (1977). Forecasting Economic Time Series, Academic Press, New York. Hannan, E. J. (1970). Multiple Time Series, John Wiley, New York. Harvey, A. C. (1980). Time Series Models, Halsted Press, New York. P

٤٥٦٧٨-

P

٩-

Montgomery, D. C., Johnson, L. A. and Gardiner, J. S. (1990). Forecasting and Time Series Analysis, 2nd ed., McGraw-Hill International Edition. ١٠Makridakis, S., Wheelwright, S. C. and McGee, V. E. (1983). Forecasting Methods and Applications, 2nd ed., P

P

P

P

John Wiley, New York. Shumway, R. H. (1988). Applied Statistical Time Series ١١Analysis, Prentice-Hall, New York. Wei, W. W. S. (1990). Time Series Analysis Univariate and ١٢Multivariate Methods, Addison Wesley. ١٣Minitab Reference Manual, Release 11 for Windows. (1998).

T

‫ﻟﻸﺳﻒ اﻟﺸﺪﻳﺪ ﻻﺗﻮﺟﺪ ﺣﺴﺐ ﻋﻠﻤﻲ ﻣﺮاﺟﻊ ﻋﺮﺑﻴﺔ ﺗﻐﻄﻲ آﻞ أو ﺟﺰء ﻣﻦ ﻣﺤﺘﻮى‬ ‫اﻟﻤﺎدة اﻟﻤﻐﻄﺎة ﻓﻲ هﺬا اﻟﻜﺘﺎب وأرﺟﻮا ﻣﻦ أي ﻃﺎﻟﺐ أو ﺑﺎﺣﺚ أو ﻣﺪرس ﻳﻌﻠﻢ ﺑﻤﺜﻞ‬ :‫هﺬا اﻟﻤﺮﺟﻊ او اﻟﻜﺘﺎب أن ﻳﺮﺳﻞ ﻟﻲ ﻣﻼﺣﻈﺔ ﻋﻠﻰ اﻟﺒﺮﻳﺪ اﻹﻟﻜﺘﺮوﻧﻲ‬ [email protected] ‫ أو‬[email protected] .‫وﺷﻜﺮا‬ T

T

T

٢٠٤

Related Documents

Time Series Good
May 2020 22
Time Series.
May 2020 35
Good Time
May 2020 21
Time Series
May 2020 27
A Good Time Pass
July 2020 7