<<<<]Ù]<Ø’ËÖخ]†Â<‹Ú êñ^’uý]<öfßjÖ]æ<íéßÚˆÖ]<؉øŠÖ]<Øé× Time Series Analysis and Forecasting
.1 .15ﻣﻘﺪﻣﺔ .2 .15ﻧﻤﺎذج ﺑﻮآﺲ-ﺟﻴﻨﻜﻨﺰ ﻟﺘﺤﻠﻴﻞ اﻟﺴﻼﺳﻞ اﻟﺰﻣﻨﻴﺔ .3 .15اﻟﻄﺮق اﻟﺘﻤﻬﻴﺪﻳﺔ ﻟﻠﺘﻌﺮف ﻋﻠﻰ ﻧﻤﻮذج اﻟﺴﻼﺳﻞ اﻟﺰﻣﻨﻴﺔ .4 .15ﻃﺮق آﻤﻴﺔ ﻟﺘﺤﺴﻴﻦ ﻧﻤﻮذج اﻟﺴﻼﺳﻞ اﻟﺰﻣﻨﻴﺔ اﻟﻤﻘﺘﺮح .5 .15اﺧﺘﻴﺎر ﻧﻤﻮذج اﻟﺴﻼﺳﻞ اﻟﺰﻣﻨﻴﺔ ﺑﺎﺳﺘﺨﺪام ﻧﻈﺎم SPSS .6 .15ﺗﺤﻠﻴﻞ ﻧﻤﺎذج اﻟﺴﻼﺳﻞ اﻟﺰﻣﻨﻴﺔ ﻏﻴﺮ اﻟﺴﺎآﻨﺔ .7 .15ﺗﺤﻠﻴﻞ اﻟﺴﻼﺳﻞ اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
516
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
517
]Ù]<Ø’ËÖخ]†Â<‹Ú êñ^’uý]<öfßjÖ]æ<íéßÚˆÖ]<؉øŠÖ]<Øé× Time Series Analysis and Forecasting .1 .15ﻣﻘﺪﻣﺔ : ﻴﻁﻠﻕ ﻋﻠﻰ ﻤﺠﻤﻭﻋﺔ ﺍﻟﻤﺸﺎﻫﺩﺍﺕ ﺍﻟﺘﻲ ﺘﻤﺜل ﻗﻴﺎﺴﺎﺕ ﻟﻅﺎﻫﺭﺓ ﻤﻌﻴﻨﺔ ﺨﻼل
ﻓﺘﺭﺍﺕ ﺯﻤﻨﻴﺔ ﻤﺤﺩﺩﺓ ﺘﻌﺒﻴﺭ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ، Time Seriesﻓﺎﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﻫﻲ ﻋﺒﺎﺭﺓ ﻋﻥ ﺴﺠل ﺘﺎﺭﻴﺨﻲ ﻟﻨﺸﺎﻁ ﻤﻌﻴﻥ ﺒﻘﻴﺎﺴﺎﺕ ﻤﺄﺨﻭﺫﺓ ﻋﻠﻰ ﻓﺘﺭﺍﺕ ﺯﻤﻨﻴﺔ ﻤﺘﺴﺎﻭﻴﺔ
ﺃﻭ ﺒﺘﻌﺒﻴﺭ ﺁﺨﺭ ﻫﻲ ﻗﻴﻡ ﻟﻤﺘﻐﻴﺭ ﻤﻌﻴﻥ ﻤﺭﺘﺒﻁ ﺒﺎﻟﺯﻤﻥ ،ﻭﻴﺠﺏ ﻓﻲ ﺠﻤﻴﻊ ﺍﻷﺤﻭﺍل ﺃﻥ
ﺘﻜﻭﻥ ﻫﺫﻩ ﺍﻟﻘﻴﺎﺴﺎﺕ ﻤﺘﻨﺎﺴﻘﺔ ﻓﻲ ﻁﺭﻴﻘﺔ ﺍﻟﻘﻴﺎﺱ ﻭﻓﻲ ﻁﺒﻴﻌﺔ ﺍﻟﻅﺎﻫﺭﺓ ﺃﻭ ﺍﻟﻨﺸﺎﻁ.
ﻭﻓﻜﺭﺓ ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﺒﺒﺴﺎﻁﺔ ﻫﻲ ﺘﻘﺩﻴﺭ ﻨﻤﻭﺫﺝ ﺭﻴﺎﻀﻲ ﻴﻤﻜﻨﻪ ﺃﻥ
ﻴﺤﺎﻜﻲ ﺘﻘﺭﻴﺒﹰﺎ ﺍﻟﺘﺩﺭﺝ ﺍﻟﺘﺎﺭﻴﺨﻲ ﻟﺘﻠﻙ ﺍﻟﻅﺎﻫﺭﺓ ﺒﺤﻴﺙ ﻴﻤﻜﻨﻪ ﺃﻥ ﻴﻘﺩﺭ ﺒﺩﻗﺔ ﻗﻴﻡ ﺍﻟﺴﻠﺴﻠﺔ
ﺍﻟﺯﻤﻨﻴﺔ ﻭﻴﻤﻜﻥ ﺍﺴﺘﺨﺩﺍﻤﻪ ﺒﺎﻟﺘﻨﺒﺅ ﺒﻘﻴﻡ ﻤﺴﺘﻘﺒﻠﻴﺔ ﻟﻬﺫﻩ ﺍﻟﻅﺎﻫﺭﺓ ،ﻭﻴﻌﺘﺒﺭ ﺃﺴﻠﻭﺏ ﺘﺤﻠﻴل ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﻓﻘﻁ ﺃﺤﺩ ﺍﻷﺴﺎﻟﻴﺏ ﺍﻟﻤﺴﺘﺨﺩﻤﺔ ﻓﻲ ﻋﻤﻠﻴﺎﺕ ﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ ،ﻓﻬﻨﺎﻙ
ﺃﺴﺎﻟﻴﺏ ﺃﺨﺭﻯ ﻗﺩ ﻴﻜﻭﻥ ﺃﺴﺎﺴﻬﺎ ﺃﻴﻀﹰﺎ ﺃﺴﻠﻭﺏ ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﺃﻭ ﺃﺴﻠﻭﺏ ﺍﻻﻨﺤﺩﺍﺭ ﺍﻟﻤﺘﻌﺩﺩ ﺃﻭ ﺃﻱ ﺃﺴﻠﻭﺏ ﺁﺨﺭ ،ﻭﺤﻴﺙ ﺃﻨﻨﺎ ﺘﻁﺭﻗﻨﺎ ﺇﻟﻰ ﺃﺴﻠﻭﺏ ﺍﻻﻨﺤﺩﺍﺭ ﻓﻲ
ﺍﻟﻔﺼل ﺍﻟﺜﺎﻨﻲ ﻋﺸﺭ ﻤﻥ ﻫﺫﺍ ﺍﻟﻜﺘﺎﺏ ﻓﺈﻨﻨﺎ ﺍﻵﻥ ﺴﻨﻭﻀﺢ ﻁﺭﻴﻘﺔ ﺍﺴﺘﺨﺩﺍﻡ ﺃﺴﻠﻭﺏ ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻓﻲ ﻫﺫﺍ ﺍﻟﻔﺼل.
ﻭﺘﻘﻭﻡ ﻁﺭﻴﻘﺔ ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻋﻠﻰ ﻓﻜﺭﺓ ﺇﻴﺠﺎﺩ ﻨﻤﻭﺫﺝ ﺭﻴﺎﻀﻲ
ﻤﻨﺎﺴﺏ ﻟﻁﺒﻴﻌﺔ ﺍﻟﺒﻴﺎﻨﺎﺕ ﺒﺤﻴﺙ ﻴﺠﻌل ﺍﻟﺒﻭﺍﻗﻲ )ﺍﻷﺨﻁﺎﺀ( ) Residualsﻭﻫﻲ ﺍﻟﻔﺭﻕ
ﺒﻴﻥ ﺍﻟﻘﻴﻡ ﺍﻟﺤﻘﻴﻘﻴﺔ ﻟﻠﺴﻠﺴﻠﺔ ﻭﺍﻟﻘﻴﻡ ﺍﻟﻤﻘﺩﺭﺓ ﺒﺎﺴﺘﺨﺩﺍﻡ ﺫﻟﻙ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺭﻴﺎﻀﻲ( ﺃﻗل ﻤﺎ
ﻴﻤﻜﻥ ﻭﻟﻴﺱ ﺒﻬﺎ ﺃﻱ ﻨﻭﻉ ﻤﻥ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺩﺍﺨﻠﻲ ﻓﻴﻤﺎ ﺒﻴﻨﻬﺎ.
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
518
ﻭﻫﻨﺎﻙ ﺨﻁﻭﺍﺕ ﻤﺤﺩﺩﺓ ﻟﺒﻨﺎﺀ ﻨﻤﻭﺫﺝ ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻴﻤﻜﻥ ﺘﻜﺭﺍﺭﻫﺎ ﻜﻤﺎ ﻨﺸﺎﺀ ﻟﺤﻴﻥ ﺇﻴﺠﺎﺩ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻷﻜﺜﺭ ﻤﻼﺌﻤﺔ ﻟﻁﺒﻴﻌﺔ ﺍﻟﺒﻴﺎﻨﺎﺕ ﻭﻴﺼﻑ ﺍﻟﺘﻐﻴﺭﺍﺕ ﻓﻲ
ﺍﻟﻅﺎﻫﺭﺓ ﻤﻭﻀﻭﻉ ﺍﻟﺩﺭﺍﺴﺔ ﺒﺄﻋﻠﻰ ﺩﻗﺔ ﻤﻤﻜﻨﺔ ،ﻭﻴﻌﺘﺒﺭ ﺍﻷﺴﻠﻭﺏ ﺍﻟﺫﻱ ﺍﻜﺘﺸﻔﻪ ﺍﻟﻌﺎﻟﻤﺎﻥ ﺒﻭﻜﺱ ﻭﺠﻴﻨﻜﻨﺯ Box and Jenkinsﺃﻫﻡ ﺍﻷﺴﺎﻟﻴﺏ ﺍﻟﻤﺴﺘﺨﺩﻤﺔ ﻟﺒﻨﺎﺀ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﻤﺨﺘﻠﻔﺔ
ﻓﻲ ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ،ﻭﻴﻌﺘﻤﺩ ﺃﺴﻠﻭﺏ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ
Box-Jenkins
Approachﻋﻠﻰ ﻤﺠﻤﻭﻋﺔ ﻤﻥ ﺍﻷﺴﺱ ﺍﻹﺤﺼﺎﺌﻴﺔ ﺍﻟﻬﺎﻤﺔ ﻭﻴﻤﻜﻥ ﺍﺴﺘﺨﺩﺍﻤﻪ ﻓﻲ ﺘﺤﻠﻴل ﻋﺩﺩ ﻜﺒﻴﺭ ﺠﺩﹰﺍ ﻤﻥ ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻟﻅﻭﺍﻫﺭ ﻓﻲ ﻤﺨﺘﻠﻑ ﺍﻟﻤﻴﺎﺩﻴﻥ ،ﻓﻬﺫﺍ
ﺍﻷﺴﻠﻭﺏ ﻴﺘﻀﻤﻥ ﻋﺩﺩ ﻜﺒﻴﺭ ﺠﺩﹰﺍ ﻤﻥ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﺭﻴﺎﻀﻴﺔ ﺍﻟﻤﻨﺎﺴﺒﺔ ﻟﺘﻤﺜﻴل ﻅﻭﺍﻫﺭ ﻜﺜﻴﺭﺓ
ﺠﺩﹰﺍ ﻭﺍﻟﺘﻲ ﻴﻤﻜﻥ ﺍﻻﺨﺘﻴﺎﺭ ﻤﻥ ﺒﻴﻨﻬﺎ ،ﺒﺎﻹﻀﺎﻓﺔ ﺇﻟﻰ ﺫﻟﻙ ﻓﻬﻨﺎﻙ ﻁﺭﻴﻘﺔ ﻤﻭﺤﺩﺓ ﻟﻠﺘﺤﻘﻕ
ﻤﻥ ﺩﻗﺔ ﻫﺫﺍ ﺍﻟﻨﻤﻭﺫﺝ ﻓﻲ ﺘﻤﺜﻴل ﺍﻟﺒﻴﺎﻨﺎﺕ ،ﻓﻴﺼﺎﺤﺏ ﻫﺫﺍ ﺍﻷﺴﻠﻭﺏ ﻋﺩﺩ ﻤﻥ ﺍﻻﺨﺘﺒﺎﺭﺍﺕ ﺍﻹﺤﺼﺎﺌﻴﺔ ﺍﻟﺘﻲ ﻴﻤﻜﻨﻬﺎ ﺃﻥ ﺘﻤﻜﻨﻨﺎ ﻤﻥ ﺍﻟﺘﻌﺭﻑ ﻋﻠﻰ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﻤﻨﺎﺴﺏ
ﻟﻠﺒﻴﺎﻨﺎﺕ ،ﻋﻠﻰ ﺍﻟﻌﻜﺱ ﻤﻥ ﺍﻟﻁﺭﻕ ﺍﻟﺘﻘﻠﻴﺩﻴﺔ ﺍﻟﻤﺘﻌﺎﺭﻑ ﻋﻠﻴﻬﺎ ﺍﻟﺘﻲ ﺘﺸﻤل ﻋﺩﺩ ﻤﺤﺩﻭﺩ
ﺠﺩﹰﺍ ﻤﻥ ﺍﻟﻨﻤﺎﺫﺝ ﻭﺍﻟﺘﻲ ﻗﺩ ﻻ ﻴﻤﻜﻨﻬﺎ ﻭﺼﻑ ﺍﻟﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﻌﻘﺩﺓ ﻓﻲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ
ﻨﺎﻫﻴﻙ ﻋﻥ ﻋﺩﻡ ﻗﺩﺭﺘﻬﺎ ﻋﻠﻰ ﺍﺴﺘﺨﺩﺍﻡ ﺍﻻﺨﺘﺒﺎﺭﺍﺕ ﺍﻹﺤﺼﺎﺌﻴﺔ ﺍﻟﻤﻼﺌﻤﺔ ﻟﻠﺘﺤﻘﻕ ﻤﻥ ﺼﺤﺔ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺫﻱ ﻴﺘﻡ ﺘﻭﻓﻴﻘﻪ.
ﻭﺘﻅل ﺩﻗﺔ ﺍﻟﺒﻴﺎﻨﺎﺕ ﺍﻟﻌﺎﻤل ﺍﻷﺴﺎﺴﻲ ﻓﻲ ﻋﺩﻡ ﻓﻬﻡ ﺃﻭ ﺘﺭﺠﻤﺔ ﺃﻭ ﺩﻗﺔ ﺍﻟﻨﺘﺎﺌﺞ ﻓﻲ
ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ،ﻓﻼ ﺒﺩ ﻤﻥ ﺍﻟﺘﺄﻜﻴﺩ ﻫﻨﺎ ﻋﻠﻰ ﻭﺠﻭﺏ ﺘﻨﺎﺴﻕ ﺍﻟﺒﻴﺎﻨﺎﺕ ﻭﻭﻀﻭﺤﻬﺎ ﻭﻭﻀﻭﺡ ﻜﻴﻔﻴﺔ ﺠﻤﻌﻬﺎ ﺃﻭ ﻗﻴﺎﺴﻬﺎ ،ﻜﻤﺎ ﺃﻨﻪ ﻟﻜﻲ ﻴﻜﻭﻥ ﺍﻟﺘﺤﻠﻴل ﻭﺍﻟﺘﻨﺒﺅ
ﻑ ﻤﻥ ﺍﻟﺒﻴﺎﻨﺎﺕ ،ﻭﺘﻌﺘﺒﺭ ﺍﻟﺴﻠﺴﻠﺔ ﺒﻘﻴﻡ ﺍﻹﺤﺼﺎﺌﻲ ﺩﻗﻴﻘﹰﺎ ﻻﺒﺩ ﻤﻥ ﺃﻥ ﻴﻜﻭﻥ ﻫﻨﺎﻙ ﻋﺩﺩ ﻜﺎ ٍ ﻋﺩﺩﻫﺎ ﻴﺘﺭﺍﻭﺡ ﻤﻥ 40ﺇﻟﻰ 50ﻗﻴﻤﺔ ﺘﻡ ﻗﻴﺎﺴﻬﺎ ﻋﻠﻰ ﻓﺘﺭﺍﺕ ﺯﻤﻨﻴﺔ ﻤﺘﺴﺎﻭﻴﺔ ﻋﺩﺩ
ﻤﻨﺎﺴﺏ ﻷﻏﺭﺍﺽ ﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ ﺒﺩﻗﺔ ﻜﺎﻓﻴﺔ ،ﻭﻓﻲ ﺒﻌﺽ ﺍﻟﺤﺎﻻﺕ ﻗﺩ ﻨﺤﺘﺎﺝ ﺇﻟﻰ
ﺇﺠﺭﺍﺀ ﺒﻌﺽ ﺍﻟﺘﻌﺩﻴﻼﺕ ﻋﻠﻰ ﺍﻟﺒﻴﺎﻨﺎﺕ ﻗﺒل ﺘﺤﻠﻴﻠﻬﺎ ﻤﺜل ﺘﻌﺩﻴل ﺍﻟﻘﻴﻡ ﺍﻟﻤﺘﻁﺭﻓﺔ ﺃﻭ
ﺍﻟﻤﻔﻘﻭﺩﺓ ﺃﻭ ﺍﺴﺘﺨﺩﺍﻡ ﺘﺤﻭﻴﻼﺕ ﻟﻭﻏﺎﺭﻴﺘﻤﻴﺔ ﺃﻭ ﻏﻴﺭﻫﺎ ﺒﺎﺴﺘﺨﺩﺍﻡ ﺍﻟﺩﻭﺍل ﺍﻟﻤﺘﺎﺤﺔ.
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
519
.2 .15ﻧﻤﺎذج ﺑﻮآﺲ-ﺟﻴﻨﻜﻨﺰ ﻟﺘﺤﻠﻴﻞ اﻟﺴﻼﺳﻞ اﻟﺰﻣﻨﻴﺔ: Box-Jenkins Approach in Time Series Analysis: ﺘﻌﺘﺒﺭ ﻤﺠﻤﻭﻋﺔ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﻌﺎﻤﺔ ﻟﻠﺘﻨﺒﺅ ﺍﻟﺘﻲ ﺍﻜﺘﺸﻔﻬﺎ ﺍﻟﻌﺎﻟﻤﺎﻥ ﺒﻭﻜﺱ ﻭﺠﻴﻨﻜﻨﺯ
Box and Jenkinsﻓﻲ ﺍﻟﻌﺎﻡ 1970ﻭﺍﻟﺘﻲ ﻴﻁﻠﻕ ﻋﻠﻴﻬﺎ ﺍﺴﻡ "ﻨﻤﺎﺫﺝ ﺍﻻﻨﺤﺩﺍﺭ ﺍﻟﺫﺍﺘﻲ ﻭﺍﻟﻤﺘﻭﺴﻁﺎﺕ ﺍﻟﻤﺘﺤﺭﻜﺔ ﺍﻟﻤﺘﻜﺎﻤﻠﺔ"
Auto-Regressive Integrated Moving
) Average Models (ARIMAﺃﻫﻡ ﺍﻷﺴﺎﻟﻴﺏ ﺍﻟﻤﺴﺘﺨﺩﻤﺔ ﻟﺒﻨﺎﺀ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﻤﺨﺘﻠﻔﺔ ﻓﻲ ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ،ﻫﺫﻩ ﺍﻷﺴﺎﻟﻴﺏ ﺘﻌﺩ ﺍﻤﺘﺩﺍﺩﹰﺍ ﻷﺴﻠﻭﺏ ﺍﻻﻨﺤﺩﺍﺭ ﺍﻟﺫﻱ ﺘﻡ ﺘﻨﺎﻭﻟﻪ
ﻓﻲ ﺍﻟﻔﺼل ﺍﻟﺜﺎﻨﻲ ﻋﺸﺭ ﻤﻥ ﻫﺫﺍ ﺍﻟﻜﺘﺎﺏ ،ﻭﺠﻤﻴﻊ ﻫﺫﻩ ﺍﻟﻨﻤﺎﺫﺝ ﺘﻌﺘﺒﺭ ﻨﻤﺎﺫﺝ ﺨﻁﻴﺔ، ﻭﻫﻲ ﻤﺘﻌﺩﺩﺓ ﺍﻟﺠﻭﺍﻨﺏ ﻭﺴﻴﺘﻡ ﺘﻭﻀﻴﺢ ﻁﺭﻴﻘﺔ ﺍﺨﺘﻴﺎﺭ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﻤﻼﺌﻡ ﻟﻠﺒﻴﺎﻨﺎﺕ ﻓﻴﻤﺎ
ﻴﻠﻲ ،ﻭﻫﻨﺎﻙ ﺜﻼﺙ ﻓﺌﺎﺕ ﻋﺎﻤﺔ ﻤﻥ ﻫﺫﻩ ﺍﻟﻨﻤﺎﺫﺝ ،ﻭﻟﺘﻭﻀﻴﺤﻬﺎ ﺴﻭﻑ ﻨﺴﺘﺨﺩﻡ ﺍﻟﺭﻤﻭﺯ ﺍﻟﺘﺎﻟﻴﺔ:
: Xtﺘﺸﻴﺭ ﺇﻟﻰ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺒﻭﺠﻪ ﻋﺎﻡ ﻭﺇﻟﻰ ﻗﻴﻤﺔ ﺍﻟﻅﺎﻫﺭﺓ ﻓﻲ ﺍﻟﻔﺘﺭﺍﺕ ﺍﻟﺯﻤﻨﻴﺔ ، t=1,2,.....,m
: φ iﺘﺸﻴﺭ ﺇﻟﻰ ﻤﻌﺎﻟﻡ ﻋﻭﺍﻤل ﺍﻻﻨﺤﺩﺍﺭ ﺍﻟﺫﺍﺘﻲ ، AR
: θ jﺘﺸﻴﺭ ﺇﻟﻰ ﻤﻌﺎﻟﻡ ﻋﻭﺍﻤل ﺍﻟﻤﺘﻭﺴﻁﺎﺕ ﺍﻟﻤﺘﺤﺭﻜﺔ ، MA : etﺘﺸﻴﺭ ﺇﻟﻰ ﺍﻷﺨﻁﺎﺀ ﻓﻲ ﺍﻟﻔﺘﺭﺍﺕ ﺍﻟﺯﻤﻨﻴﺔ ، t=1,2,.....,m
: Cﺘﺸﻴﺭ ﺇﻟﻰ ﺜﺎﺒﺕ ﺍﻟﻨﻤﻭﺫﺝ.
ﻭﺒﺎﻟﺘﺎﻟﻲ ﻴﻤﻜﻥ ﺘﻭﻀﻴﺢ ﻓﺌﺎﺕ ﻨﻤﺎﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﻟﻠﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻜﻤﺎ ﻴﻠﻲ: .1ﻨﻤﺎﺫﺝ ﺍﻻﻨﺤﺩﺍﺭ ﺍﻟﺫﺍﺘﻲ : Autoregressive ﻴﻤﻜﻥ ﻜﺘﺎﺒﺔ ﻨﻤﻭﺫﺝ ﺍﻻﻨﺤﺩﺍﺭ ﺍﻟﺫﺍﺘﻲ ﻤﻥ ﺍﻟﺩﺭﺠﺔ ﺍﻷﻭﻟﻰ ) AR(1ﻋﻠﻰ ﺍﻟﺼﻭﺭﺓ
ﺍﻟﺘﺎﻟﻴﺔ :
X t =C +φX t −1 +et
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
520
ﻭﻴﻤﻜﻥ ﺘﻌﻤﻴﻡ ﻫﺫﺍ ﺍﻟﻨﻤﻭﺫﺝ ﻟﻴﺼﺒﺢ ﻤﻥ ﺍﻟﺩﺭﺠﺔ pﻟﻴﺼﺒﺢ ) AR(pﻭﻴﺄﺨﺫ ﺍﻟﺼﻭﺭﺓ : X t =C +φ1 X t −1 +φ 2 X t −2 +.... +φ p X t − p + et
.2ﻨﻤﺎﺫﺝ ﺍﻟﻤﺘﻭﺴﻁﺎﺕ ﺍﻟﻤﺘﺤﺭﻜﺔ : Moving Average ﻴﻤﻜﻥ ﻜﺘﺎﺒﺔ ﻨﻤﻭﺫﺝ ﺍﻟﻤﺘﻭﺴﻁﺎﺕ ﺍﻟﻤﺘﺤﺭﻜﺔ ﻤﻥ ﺍﻟﺩﺭﺠﺔ ﺍﻷﻭﻟﻰ ) MA(1ﻋﻠﻰ
ﺍﻟﺼﻭﺭﺓ ﺍﻟﺘﺎﻟﻴﺔ :
X t =C +et +θet −1
ﻭﻴﻤﻜﻥ ﺘﻌﻤﻴﻡ ﻫﺫﺍ ﺍﻟﻨﻤﻭﺫﺝ ﻟﻴﺼﺒﺢ ﻤﻥ ﺍﻟﺩﺭﺠﺔ qﻟﻴﺼﺒﺢ ) MA(qﻭﻴﺄﺨﺫ
ﺍﻟﺼﻭﺭﺓ :
X t =C +et +θ 1et −1 +θ 2 et −2 + ..... +θ q et −q
.3ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﻤﺨﺘﻠﻁﺔ : Mixed Model ﻭﻫﺫﺍ ﺍﻟﻨﻤﻭﺫﺝ ﻋﺒﺎﺭﺓ ﻋﻥ ﺨﻠﻴﻁ ﻤﻥ ﺍﻟﻨﻤﻭﺫﺠﻴﻥ ﺍﻟﺴﺎﺒﻘﻴﻥ ،ﻭﻴﻤﻜﻥ ﻜﺘﺎﺒﺔ ﺍﻟﻨﻤﻭﺫﺝ ) ARMA (p,qﻋﻠﻰ ﺍﻟﺼﻭﺭﺓ ﺍﻟﻌﺎﻤﺔ ﺍﻟﺘﺎﻟﻴﺔ : X t =C +φ1 X t −1 +φ 2 X t −2 +.... +φ p X t − p + et +θ 1et −1 +θ 2 et − 2 + ..... +θ q et − q
ﻭﻫﺫﺍ ﺍﻟﻨﻤﻭﺫﺝ ﻴﻤﻜﻥ ﺘﻌﻤﻴﻤﻪ ﺃﻴﻀﹰﺎ ﻟﻴﺸﻤل ﺃﺨﺫ ﺍﻟﻔﺭﻭﻕ Differencingﻟﻘﻴﻡ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﻟﻠﺤﺼﻭل ﻋﻠﻰ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﻤﺨﺘﻠﻁ ﺍﻟﻤﺘﻜﺎﻤل ) ARIMA(p,d,qﺤﻴﺙ ﻴﺸﻴﺭ ﺍﻟﺤﺭﻑ Iﺇﻟﻰ ﺘﻌﺒﻴﺭ ﺍﻟﻤﺘﻜﺎﻤل Integratedﻭﻴﺸﻴﺭ ﺍﻟﺩﻟﻴل dﺇﻟﻰ ﻋﺩﺩ ﻤﺭﺍﺕ
ﺃﺨﺫ ﺍﻟﻔﺭﻭﻕ ﻟﻘﻴﻡ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ،ﻓﻌﻠﻰ ﺴﺒﻴل ﺍﻟﻤﺜﺎل ﻋﻨﺩﻤﺎ ﺘﻜﻭﻥ d=1ﻓﺈﻨﻪ ﻴﺘﻡ ﺘﻨﻔﻴﺫ ﺍﻟﻔﺭﻭﻕ ﻤﻥ ﺍﻟﺩﺭﺠﺔ ﺍﻷﻭﻟﻰ ،ﻭﻋﻨﺩﻤﺎ ﺘﻜﻭﻥ d=2ﻓﺈﻥ ﻋﻤﻠﻴﺔ ﺍﻟﻔﺭﻭﻕ ﻟﻠﻘﻴﻡ ﻗﺩ ﺃﺠﺭﻴﺕ ﻤﺭﺘﻴﻥ ،ﻭﻫﻜﺫﺍ..
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
521
.3 .15اﻟﻄﺮق اﻟﺘﻤﻬﻴﺪﻳﺔ ﻟﻠﺘﻌﺮف ﻋﻠﻰ ﻧﻤﻮذج اﻟﺴﻼﺳﻞ اﻟﺰﻣﻨﻴﺔ: Preliminary Model Identification Procedures: ﻟﻠﺘﻌﺭﻑ ﻋﻠﻰ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﻤﻨﺎﺴﺏ ﻟﺒﻴﺎﻨﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﻻﺒﺩ ﻜﺨﻁﻭﺓ ﺃﻭﻟﻰ ﻤﻥ ﺘﺤﻠﻴل ﺍﻟﺒﻴﺎﻨﺎﺕ ﻤﺒﺩﺌﻴ ﹰﺎ ﺒﺎﺴﺘﺨﺩﺍﻡ ﻨﻤﻭﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ Box-Jenkins Analysis
ﺒﺤﻴﺙ ﻴﺸﻤل ﻫﺫﺍ ﺍﻟﺘﺤﻠﻴل ﺭﺴﻤﹰﺎ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺍﻷﺼﻠﻴﺔ ،ﻭﻴﺠﺏ ﺘﻌﺩﻴل ﺍﻟﺒﻴﺎﻨﺎﺕ ﺍﻷﺼﻠﻴﺔ ﺩﺍﺌﻤﹰﺎ ﻟﻠﺤﺼﻭل ﻋﻠﻰ ﺴﻠﺴﻠﺔ ﺴﺎﻜﻨﺔ a stationary seriesﻤﺎ ﻟﻡ ﺘﻜﻥ ﻫﻲ
ﻼ ﺴﻠﺴﻠﺔ ﺴﺎﻜﻨﺔ )ﻭﻫﻲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺘﻲ ﺘﺘﻐﻴﺭ ﻗﻴﻤﻬﺎ ﻓﻲ ﺤﺩﻭﺩ ﺜﺎﺒﺘﺔ ﻁﻭﺍل ﺍﻟﻔﺘﺭﺓ ﺃﺼ ﹰ
ﺍﻟﺯﻤﻨﻴﺔ ،ﺃﻱ ﺤﺩﻭﺩ ﺘﻐﻴﺭ ﻗﻴﻡ ﺍﻟﺴﻠﺴﻠﺔ ﻻ ﻴﻌﺘﻤﺩ ﻋﻠﻰ ﺍﻟﺯﻤﻥ ﻭﻴﻤﻜﻥ ﺍﻟﺘﻌﺭﻑ ﻋﻠﻴﻬﺎ
ﺒﻭﺍﺴﻁﺔ ﺍﻟﺭﺴﻡ ﻓﻘﻁ( ،ﻜﺫﻟﻙ ﻓﺈﻥ ﺃﻱ ﺍﺘﺠﺎﻩ ﻋﺎﻡ ﻗﺩ ﻴﻅﻬﺭ ﻋﻠﻰ ﻁﻭل ﺍﻟﺴﻠﺴﻠﺔ ﻴﻤﻜﻥ ﻤﻌﺎﻟﺠﺘﻪ ﻓﻲ ﺍﻟﻨﻤﻭﺫﺝ ﺒﺎﺴﺘﺨﺩﺍﻡ ﻁﺭﻴﻘﺔ ﺍﻟﻔﺭﻭﻕ ﺍﻟﻨﻅﺎﻤﻴﺔ ،regular differencesﻭﻫﻲ
ﻋﺒﺎﺭﺓ ﻋﻥ ﻋﻤﻠﻴﺔ ﻴﺘﻡ ﻓﻴﻬﺎ ﺤﺴﺎﺏ ﺍﻟﻔﺭﻭﻕ ﺒﻴﻥ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺘﺘﺎﻟﻴﺔ ﻟﺘﻨﺘﺞ ﺴﻠﺴﻠﺔ ﻤﺴﺘﺨﻠﺼﹰﺎ ﻤﻨﻬﺎ ﺃﺜﺭ ﺍﻻﺘﺠﺎﻩ ﺍﻟﻌﺎﻡ ،ﻭﺇﺫﺍ ﻟﻡ ﻴﺤﻘﻕ ﺘﻁﺒﻴﻕ ﻁﺭﻴﻘﺔ ﺍﻟﻔﺭﻭﻕ ﺴﻜﻭﻨﹰﺎ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ
ﻤﻥ ﺍﻟﻤﺭﺓ ﺍﻷﻭﻟﻰ ﻓﺈﻨﻪ ﻴﻤﻜﻥ ﺘﻜﺭﺍﺭ ﻫﺫﻩ ﺍﻟﻌﻤﻠﻴﺔ ﻋﻠﻰ ﺍﻟﺭﻏﻡ ﻤﻥ ﺃﻨﻪ ﻨﺎﺩﺭﹰﺍ ﻤﺎ ﺘﺠﺩ
ﺴﻠﺴﻠﺔ ﻻ ﺘﺤﻘﻕ ﺍﻟﺴﻜﻭﻥ ﺒﺘﻁﺒﻴﻕ ﻁﺭﻴﻘﺔ ﺍﻟﻔﺭﻭﻕ ﻤﺭﺓ ﻭﺍﺤﺩﺓ ﺃﻭ ﻤﺭﺘﻴﻥ ﻋﻠﻰ ﺍﻷﻜﺜﺭ،
ﻭﺇﺫﺍ ﺍﺴﺘﻤﺭ ﻅﻬﻭﺭ ﻋﺩﻡ ﺍﻻﻨﺘﻅﺎﻡ ﻓﻲ ﺴﻠﺴﻠﺔ ﺍﻟﻔﺭﻭﻕ ﻓﺈﻨﻪ ﻴﻤﻜﻥ ﺘﻁﺒﻴﻕ ﺃﺤﺩ ﺍﻟﺘﺤﻭﻴﻼﺕ ﻤﺜل ﻟﻭﻏﺎﺭﻴﺘﻡ ﻗﻴﻡ ﺍﻟﺴﻠﺴﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺃﻭ ﻤﻘﻠﻭﺏ ﻫﺫﻩ ﺍﻟﻘﻴﻡ ﻟﺘﺤﻘﻴﻕ ﺍﻟﺴﻜﻭﻥ ﻟﻬﺎ ﻟﻜﻲ ﻴﺘﺒﻘﻰ
ﻓﻲ ﺍﻟﻨﻬﺎﻴﺔ ﺴﻠﺴﻠﺔ ﺒﻭﺍﻗﻲ Residualsﺘﻅﻬﺭ ﻗﻴﻤﻬﺎ ﻓﻲ ﺍﻟﺭﺴﻡ ﻗﺭﻴﺒﺔ ﻤﻥ ﺍﻟﺼﻔﺭ ﻭﻻ
ﺘﻌﻜﺱ ﺃﻱ ﺘﻐﻴﺭ ﻨﻅﺎﻤﻲ ،ﻫﺫﻩ ﺍﻟﺴﻠﺴﻠﺔ ﻟﻸﺨﻁﺎﺀ ﻴﻁﻠﻕ ﻋﻠﻴﻬﺎ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﻌﺸﻭﺍﺌﻴﺔ ﺍﻟﻨﻘﻴﺔ
Pure Random Seriesﺃﻭ ﺴﻠﺴﻠﺔ ﺍﻟﺸﻭﺍﺌﺏ ﺍﻟﺒﻴﻀﺎﺀ .White Noise
ﻜﻤﺎ ﺃﻨﻪ ﺇﺫﺍ ﺤﺩﺙ ﻭﻅﻬﺭﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺍﻷﺼﻠﻴﺔ ﺒﻼ ﺃﻱ ﺍﺘﺠﺎﻩ ﻋﺎﻡ ﺃﻭ ﺃﻱ ﺃﺜﺭ ﻟﻠﻤﻭﺴﻡ ﻭﻅﻬﺭﺕ ﺴﻠﺴﻠﺔ ﺍﻟﺒﻭﺍﻗﻲ Residualsﻗﻴﻡ ﻗﺭﻴﺒﺔ ﺠﺩﹰﺍ ﻤﻥ ﺍﻟﺼﻔﺭ ﻓﻲ
ﺤﺩﻭﺩ 95%ﻓﺘﺭﺓ ﺜﻘﺔ ﻭﺒﺩﻭﻥ ﻅﻬﻭﺭ ﺃﻱ ﺘﻐﻴﺭﺍﺕ ﻤﻨﺘﻅﻤﺔ ﻓﺈﻥ ﺍﻟﺴﻠﺴﺔ ﻓﻲ ﻫﺫﻩ ﺍﻟﺤﺎﻟﺔ
ﺃﻴﻀﹰﺎ ﻫﻲ ﺴﻠﺴﻠﺔ ﻨﻘﻴﺔ ﺃﻭ ﺸﻭﺍﺌﺏ ﺒﻴﻀﺎﺀ ﻭﻟﻥ ﺘﺘﻤﻜﻥ ﺍﻟﻁﺭﻕ ﺍﻹﺤﺼﺎﺌﻴﺔ ﻤﻥ ﺘﺤﻠﻴﻠﻬﺎ.
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
522
ﺇﺫﺍ ﻜﺎﻨﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ )ﺃﻭ ﺃﺼﺒﺤﺕ( ﺴﻠﺴﻠﺔ ﺴﺎﻜﻨﺔ stationaryﻓﺈﻨﻪ ﻴﻤﻜﻥ ﺍﻟﺘﻌﺭﻑ ﺍﻵﻥ ﻋﻠﻰ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﻤﺒﺩﺌﻲ ﺍﻟﻤﻤﺜل ﻟﻠﺒﻴﺎﻨﺎﺕ ،ﻭﺘﻌﺘﺒﺭ ﺍﻟﺜﻼﺙ ﻨﻤﺎﺫﺝ ﺍﻟﺘﻲ ﺘﻡ
ﺫﻜﺭﻫﺎ ﻭﻫﻲ ARﻭ MAﻭﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﻤﺸﺘﺭﻙ ARMAﺇﻀﺎﻓﺔ ﺇﻟﻰ ﻋﻤﻠﻴﺔ ﺍﻟﻔﺭﻭﻕ ﺍﻟﻨﻅﺎﻤﻴﺔ ) Regular Differences (RDﻨﻤﺎﺫﺝ ﻤﺒﺩﺌﻴﺔ ﻤﻤﻜﻨﺔ ،ﻫﺫﻩ ﺍﻟﻨﻤﺎﺫﺝ ﺘﺠﺘﻤﻊ ﻤﻌﹰﺎ ﻟﺘﻜﻭﻥ ﺍﻷﺩﻭﺍﺕ ﺍﻟﺭﺌﻴﺴﻴﺔ ﻟﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ،ﻭﺇﺫﺍ ﺘﻡ ﺘﻁﺒﻴﻕ ﻁﺭﻴﻘﺔ ﺍﻟﻔﺭﻭﻕ
ﺍﻟﻨﻅﺎﻤﻴﺔ RDﻤﻊ ﺃﻱ ﻤﻥ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﺴﺎﺒﻘﺔ ﻓﺈﻨﻪ ﻴﻨﺘﺞ ﻓﻲ ﻫﺫﻩ ﺍﻟﺤﺎﻟﺔ ﻨﻤﻭﺫﺝ ،ARIMA ﻭﻴﻤﻜﻥ ﺘﺤﺩﻴﺩ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﻤﺒﺩﺌﻲ ﻟﻠﺴﻠﺴﻠﺔ ﺒﺩﻗﺔ ﻋﺎﻟﻴﺔ ﻤﻥ ﺨﻼل ﺭﺴﻡ ﺩﺍﻟﺘﻲ ﺍﻟﺘﺭﺍﺒﻁ
ﺍﻟﺫﺍﺘﻲ) Autocorrelation Function (ACFﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ
Partial
) ،Autocorrelation Function (PACFﻓﻤﻌﺭﻭﻑ ﺃﻥ ﻤﻥ ﺨﺼﺎﺌﺹ ﻨﻤﻭﺫﺝ
ﺍﻻﻨﺤﺩﺍﺭ ﺍﻟﺫﺍﺘﻲ ) AR(pﺃﻥ ﺩﺍﻟﺔ ACFﺘﺘﻼﺸﻰ ﺘﺩﺭﻴﺠﻴﹰﺎ ﻭﺩﺍﻟﺔ PACFﺘﺘﻭﻗﻑ ﻋﻨﺩ
ﺍﻟﻔﺘﺭﺓ ، pﺒﻴﻨﻤﺎ ﻓﻲ ﻨﻤﻭﺫﺝ ﺍﻟﻤﺘﻭﺴﻁﺎﺕ ﺍﻟﻤﺘﺤﺭﻜﺔ ) MA(qﺘﺘﻼﺸﻰ ﺩﺍﻟﺔ PACF
ﺘﺩﺭﻴﺠﻴﹰﺎ ﻭ ﺘﺘﻭﻗﻑ ﺩﺍﻟﺔ ACFﻋﻨﺩ ﺍﻟﻔﺘﺭﺓ . q
ﻭﺒﺎﻹﻀﺎﻓﺔ ﺇﻟﻰ ﻭﺠﻭﺩ ﺍﻻﺘﺠﺎﻩ ﺍﻟﻌﺎﻡ ﻓﺈﻥ ﻫﻨﺎﻙ ﺍﻟﻜﺜﻴﺭ ﻤﻥ ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﺍﻟﺴﺎﻜﻨﺔ stationary seriesﺘﻅﻬﺭ ﺘﻐﻴﺭﺍﺕ ﻤﻭﺴﻤﻴﺔ ،ﻓﻴﻜﻭﻥ ﻫﻨﺎﻙ ﺘﻭﺠﻪ ﻟﻤﺠﻤﻭﻋﺔ
ﻤﻥ ﺍﻟﺫﺒﺫﺒﺎﺕ ﻓﻲ ﻋﺩﺩ ﻤﺘﺘﺎﻟﻲ ﻭﻤﺤﺩﻭﺩ ﻤﻥ ﺍﻟﻘﻴﻡ ﺃﻥ ﺘﻌﻴﺩ ﻨﻔﺴﻬﺎ ﻓﻲ ﺍﻟﻤﻭﺍﺴﻡ ﺍﻟﻤﺘﺸﺎﺒﻬﺔ،
ﻭﻴﻤﻜﻥ ﺃﺤﻴﺎﻨﹰﺎ ﺃﻥ ﺘﻅﻬﺭ ﺍﻟﺘﻐﻴﺭﺍﺕ ﺍﻟﻤﻭﺴﻤﻴﺔ ﺍﺘﺠﺎﻫﹰﺎ ﻟﺘﻐﻴﺭ ﺇﻀﺎﻓﻲ ﻓﻴﺘﻐﻴﺭ ﻤﺩﻯ ﺘﻐﻴﺭﻫﺎ
ﻜﻠﻤﺎ ﺘﻘﺩﻡ ﺍﻟﺯﻤﻥ ،ﻭﻫﻨﺎ ﺒﺎﻟﻤﺜل ﻴﻤﻜﻥ ﺘﻁﺒﻴﻕ ﻁﺭﻴﻘﺔ ﺍﻟﻔﺭﻭﻕ ﺍﻟﻤﻭﺴﻤﻴﺔ Seasonal
) Differencing (SDﻹﺯﺍﻟﺔ ﻋﺩﻡ ﺍﻟﺴﻜﻭﻥ ﺍﻟﻤﻭﺴﻤﻲ ﻤﻥ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺍﻟﻤﻭﺴﻤﻴﺔ،
ﻭﻜﻤﺎ ﺃﻨﻪ ﻴﻤﻜﻥ ﺍﺴﺘﺨﺩﺍﻡ ﺃﺴﻠﻭﺒﻲ ﺍﻻﻨﺤﺩﺍﺭ ﺍﻟﺫﺍﺘﻲ ARﻭﺍﻟﻤﺘﻭﺴﻁﺎﺕ ﺍﻟﻤﺘﺤﺭﻜﺔ MA
ﻜﺄﺩﻭﺍﺕ ﺭﺌﻴﺴﻴﺔ ﻟﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﺒﺼﻔﺔ ﻋﺎﻤﺔ ﻓﺈﻥ ﻫﺫﻴﻥ ﺍﻷﺴﻠﻭﺒﻴﻥ ﻴﻤﻜﻥ
ﺍﺴﺘﺨﺩﺍﻤﻬﻤﺎ ﺃﻴﻀﹰﺎ ﻟﻠﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﺍﻟﻤﻭﺴﻤﻴﺔ ﺒﺎﺴﺘﺨﺩﺍﻡ "ﻤﻌﺎﻟﻡ ﻨﻤﻭﺫﺝ ﺍﻻﻨﺤﺩﺍﺭ ﺍﻟﺫﺍﺘﻲ
ﺍﻟﻤﻭﺴﻤﻲ" ) Seasonal Autoregressive Parameters (SARﻭ"ﻤﻌﺎﻟﻡ ﺍﻷﻭﺴﺎﻁ ﺍﻟﻤﺘﺤﺭﻜﺔ ﺍﻟﻤﻭﺴﻤﻴﺔ" ). Seasonal Moving Average Parameters (SMA
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
523
ﻭﺘﺘﻀﺢ ﺍﻟﺤﺎﺠﺔ ﺇﻟﻰ ﻤﻌﺎﻟﻡ ﻜل ﻤﻥ ﺍﻻﻨﺤﺩﺍﺭ ﺍﻟﺫﺍﺘﻲ ﺍﻟﻤﻭﺴﻤﻲ )(SAR ﻭﺍﻷﻭﺴﺎﻁ ﺍﻟﻤﺘﺤﺭﻜﺔ ﺍﻟﻤﻭﺴﻤﻴﺔ ) (SMAﻋﻨﺩ ﻓﺤﺹ ﻨﻤﻁ ﺩﺍﻟﺔ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ACF
ﻭﻨﻤﻁ ﺩﺍﻟﺔ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ PACFﻟﺴﻠﺴﻠﺔ ﺯﻤﻨﻴﺔ ﺴﺎﻜﻨﺔ stationary series
ﻋﻨﺩ ﻓﺘﺭﺍﺕ ﺘﺸﻜل ﻤﻀﺎﻋﻔﺎﺕ ﻟﻌﺩﺩ ﺍﻟﻭﺤﺩﺍﺕ ﺍﻟﺯﻤﻨﻴﺔ ﻓﻲ ﺍﻟﻤﻭﺴﻡ ،ﻫﺫﻩ ﺍﻟﻤﻌﺎﻟﻡ ﺴﻭﻑ
ﺘﻜﻭﻥ ﺫﺍﺕ ﺃﻫﻤﻴﺔ ﻋﻨﺩﻤﺎ ﺘﻜﻭﻥ ﺩﺍﻟﺔ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ACFﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ
PACFﻋﻨﺩ ﺍﻟﻔﺘﺭﺍﺕ sﻭ 2sﻭ ....ﺍﻟﺦ ﻏﻴﺭ ﻤﺴﺎﻭﻴﺔ ﻟﻠﺼﻔﺭ ﻭﺘﻌﻜﺱ ﻨﻤﻁﹰﺎ ﻴﺭﺘﺒﻁ ﺒﺎﻷﻨﻤﺎﻁ ﺍﻟﻨﻅﺭﻴﺔ ﻟﺘﻠﻙ ﺍﻟﻨﻤﺎﺫﺝ ،ﻭﻨﺤﺘﺎﺝ ﺇﻟﻰ ﺘﻁﺒﻴﻕ ﻁﺭﻴﻘﺔ ﺍﻟﻔﺭﻭﻕ ﺍﻟﻤﻭﺴﻤﻴﺔ ﻓﻲ ﻫﺫﻩ
ﺍﻟﻨﻤﺎﺫﺝ ﺇﺫﺍ ﺘﺒﻴﻥ ﺃﻥ ﺩﺍﻟﺔ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﻋﻨﺩ ﺍﻟﻔﺘﺭﺍﺕ ﺍﻟﻤﻭﺴﻤﻴﺔ ﻻ ﺘﺘﻼﺸﻰ ﺒﺴﺭﻋﺔ. ﻭﻴﻤﻜﻥ ﺘﻭﻀﻴﺢ ﺨﻁﻭﺍﺕ ﺒﻨﺎﺀ ﻨﻤﺎﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﺒﺎﻟﺭﺴﻡ ﺍﻟﺘﻭﻀﻴﺤﻲ ﻓﻲ
ﺍﻟﺸﻜل 1-15ﺃﺩﻨﺎﻩ ،ﻭﻓﻲ ﻫﺫﺍ ﺍﻟﺸﻜل ﻻﺒﺩ ﻤﻥ ﺍﻟﺫﻜﺭ ﺃﻥ ﺘﺒﺎﻴﻥ ﺍﻷﺨﻁﺎﺀ ﻓﻲ ﻨﻤﻭﺫﺝ
ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﻴﺠﺏ ﺃﻥ ﻴﻜﻭﻥ ﺜﺎﺒﺕ ،ﻭﻫﺫﺍ ﻴﻌﻨﻲ ﺃﻥ ﺍﻟﺘﺒﺎﻴﻥ ﻓﻲ ﻜل ﻤﺠﻤﻭﻋﺎﺕ ﺍﻟﺒﻴﺎﻨﺎﺕ ﻴﺠﺏ ﺃﻥ ﻴﻜﻭﻥ ﻤﺘﺴﺎﻭﻱ ﻭﻻ ﻴﻌﺘﻤﺩ ﻋﻠﻰ ﺍﻟﻔﺘﺭﺓ ﺍﻟﺯﻤﻨﻴﺔ ،ﻓﺈﺫﺍ ﺘﻡ ﺍﺨﺘﺭﺍﻕ ﻫﺫﺍ
ﺍﻟﺸﺭﻁ ﻓﻼﺒﺩ ﻤﻥ ﻤﻌﺎﻟﺠﺔ ﺍﻟﺒﻴﺎﻨﺎﺕ ﺒﺈﺠﺭﺍﺀ ﺘﺤﻭﻴﻠﺔ ﻻﺴﺘﻘﺭﺍﺭ ﺍﻟﺘﺒﺎﻴﻥ ،ﻭﻴﺠﺏ ﺍﻟﺘﺄﻜﻴﺩ
ﻋﻠﻰ ﺃﻨﻪ ﻻ ﻴﻭﺠﺩ ﻨﻤﻁ ﻤﺤﺩﺩ ﻓﻲ ﻫﺫﻩ ﺍﻷﺨﻁﺎﺀ ،ﻜﻤﺎ ﻴﺠﺏ ﺃﻻ ﺘﻅﻬﺭ ﻗﻴﻡ ﻤﺘﻁﺭﻓﺔ ﺃﻭ
ﺸﺎﺫﺓ ﺃﻭ ﺍﻨﺤﺭﺍﻑ ﺨﻁﻴﺭ ﻓﻲ ﺍﻟﺴﻠﺴﻠﺔ ﻴﺅﺩﻱ ﺇﻟﻰ ﻗﺴﻤﺔ ﺍﻟﺴﻠﺴﻠﺔ ﺇﻟﻰ ﺃﺠﺯﺍﺀ ﻏﻴﺭ
ﻤﺘﺠﺎﻨﺴﺔ ﺍﻷﻤﺭ ﺍﻟﺫﻱ ﻗﺩ ﻴﻌﻤل ﻋﻠﻰ ﺘﺸﻭﻴﻪ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺫﻱ ﻴﺘﻡ ﺍﺸﺘﻘﺎﻗﻪ ،ﻜﻤﺎ ﺃﻨﻪ ﻻ
ﻴﺠﺏ ﺃﻥ ﻴﻅﻬﺭ ﻓﻲ ﺍﻷﺨﻁﺎﺀ ﺃﻱ ﻨﻤﻁ ﻴﺩل ﻋﻠﻰ ﻭﺠﻭﺩ ﺫﺒﺫﺒﺎﺕ ﻤﻭﺴﻤﻴﺔ ﺒﻬﺎ ،ﻭﺍﻟﺴﺒﺏ
ﻓﻲ ﻜل ﻫﺫﻩ ﺍﻟﺸﺭﻭﻁ ﺃﻨﻪ ﺍﺨﺘﺭﺍﻗﻬﺎ ﻗﺩ ﻴﺅﺩﻱ ﺇﻟﻰ ﻅﻬﻭﺭ ﺘﻘﺩﻴﺭ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ACF
ﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ PACFﺒﺸﻜل ﻗﺩ ﻴﻔﻬﻡ ﺒﻁﺭﻴﻘﺔ ﺨﺎﻁﺌﺔ ﺃﻥ ﻫﺫﻩ ﺍﻟﺒﻴﺎﻨﺎﺕ
ﺘﺘﻁﺎﺒﻕ ﻤﻊ ﻨﻤﻭﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﺍﻟﻤﺘﻜﺎﻤل ،ARIMAﻜﻤﺎ ﺃﻥ ﻭﺠﻭﺩ ﻫﺫﻩ
ﺍﻻﺨﺘﺭﺍﻗﺎﺕ ﻗﺩ ﻴﺅﺩﻱ ﺇﻟﻰ ﺘﺸﻭﻴﺵ ﺃﻭ ﺇﺨﻔﺎﺀ ﺘﺭﻜﻴﺏ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺤﻘﻴﻘﻲ ،ﻓﻭﺠﻭﺩ ﻗﻴﻤﺔ
ﻼ ﺴﻭﻑ ﻴﺅﺩﻱ ﺇﻟﻰ ﺇﺨﻔﺎﺀ ﺍﻟﻤﻼﻤﺢ ﺍﻟﻜﺎﻤﻨﺔ ﻟﻠﻨﻤﻭﺫﺝ ﺍﻟﺤﻘﻴﻘﻲ ﺸﺎﺫﺓ ﻭﺍﺤﺩﺓ ﻓﻘﻁ ﻤﺜ ﹰ ﻭﺒﺭﻭﺯ ﺘﺄﺜﻴﺭ ﻫﺫﻩ ﺍﻟﻘﻴﻤﺔ ﻓﻘﻁ.
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
524
ﺸﻜل : 1-15ﺨﻁﻭﺍﺕ ﺒﻨﺎﺀ ﻨﻤﺎﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﺃﺭﺴﻡ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ
ﻻ
ﻁﺒﻕ ﺃﺤﺩ
ﻫل ﺍﻟﺘﺒﺎﻴﻥ ﻤﺴﺘﻘﺭ ؟
ﺍﻟﺘﺤﻭﻴﻼﺕ ﻨﻌﻡ ﺍﺤﺼل ﻋﻠﻰ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ
ﻁﺒﻕ ﻁﺭﻴﻘﺔ ﻓﺭﻭﻕ
ﻻ
ﻋﺎﺩﻴﺔ ﻭﻤﻭﺴﻤﻴﺔ
ﻫل ﺍﻟﻤﺘﻭﺴﻁ ﻤﺴﺘﻘﺭ ؟ ﻨﻌﻡ
ﺍﺨﺘﻴﺎﺭ ﺍﻟﻨﻤﻭﺫﺝ
ﻗﺩﺭ ﻤﻌﺎﻟﻡ ﺍﻟﻨﻤﻭﺫﺝ
ﻫل ﺍﻟﺒﻭﺍﻗﻲ ﻏﻴﺭ
ﻻ
ﻋﺩل ﺍﻟﻨﻤﻭﺫﺝ
ﻤﺘﺭﺍﺒﻁﺔ ؟
ﻨﻌﻡ
ﺍﺴﺘﺨﺩﻡ ﺍﻟﻨﻤﻭﺫﺝ ﻓﻲ ﺍﻟﺘﻨﺒﺅ
ﻨﻌﻡ
ﻫل ﺍﻟﻤﻌﺎﻟﻡ ﻤﻌﻨﻭﻴﺔ ﻭﻏﻴﺭ ﻤﺘﺭﺍﺒﻁﺔ ؟
ﻻ
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
525
.4 .15ﻃﺮق آﻤﻴﺔ ﻟﺘﺤﺴﻴﻦ ﻧﻤﻮذج اﻟﺴﻼﺳﻞ اﻟﺰﻣﻨﻴﺔ اﻟﻤﻘﺘﺮح : Improved Quantitative Identification Method: ﺜﻤﺔ ﻁﺭﻴﻘﺔ ﻤﺤﺴﻨﺔ ﻤﺘﺎﺤﺔ ﺍﻵﻥ ﻹﺠﺭﺍﺀ ﺘﺤﻠﻴل ﻟﻨﻤﺎﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﺍﻟﻤﺘﻜﺎﻤﻠﺔ ARIMAﻭﻤﻥ ﺸﺄﻨﻬﺎ ﺍﻟﺘﺨﻔﻴﻑ ﻤﻥ ﻤﺘﻁﻠﺒﺎﺕ ﺍﻟﻤﻨﻅﻭﺭ ﺍﻟﻤﻭﺴﻤﻲ ﻓﻲ ﺘﻘﻴﻴﻡ
ﻨﻤﻁ ﺩﺍﻟﺘﻲ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ Autocorrelationﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ Partial
Autocorrelationﻟﻸﺨﻁﺎﺀ ﻭﺍﻟﻠﺘﻴﻥ ﻗﺩ ﺘﻜﻭﻨﺎ ﻏﺎﻤﻀﺘﻴﻥ ﺒﻬﺩﻑ ﺘﺤﺩﻴﺩ ﻨﻤﻭﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﺍﻟﻤﻨﺎﺴﺏ ﻭﺍﻟﺫﻱ ﻴﻤﻜﻥ ﺍﻻﻋﺘﻤﺎﺩ ﻋﻠﻴﻪ ﻓﻲ ﺇﺠﺭﺍﺀ ﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ.
) : ARMA(1,0ﻫﺫﺍ ﺍﻟﻨﻤﻭﺫﺝ ﻻﺒﺩ ﺃﻥ ﻴﻜﻭﻥ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻷﻭل ﻴﺨﻀﻊ ﻟﻼﺨﺘﺒﺎﺭ
ﻜﻨﻤﻭﺫﺝ ﻤﻨﺎﺴﺏ ﻷﻱ ﺴﻠﺴﻠﺔ ﺯﻤﻨﻴﺔ ﺴﺎﻜﻨﺔ ،ﻭﻫﻭ ﻨﻤﻭﺫﺝ ﺒﺴﻴﻁ ﻭﻴﺤﺘﻭﻱ ﻋﻠﻰ ﻤﻌﻠﻤﺔ ﻼ ﻭﺤﻴﺩﺓ ﻟﻼﻨﺤﺩﺍﺭ ﺍﻟﺫﺍﺘﻲ ﻋﻨﺩ ﺍﻟﻔﺘﺭﺓ ﺍﻷﻭﻟﻰ ،ﻭﻴﺠﺏ ﺍﺨﺘﺒﺎﺭ ﻤﻌﻨﻭﻴﺔ ﻨﻤﻁ ﻤﻌﺎﻤﻼﺕ ﻜ ﹰ
ﻤﻥ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ ﻭﺒﺎﻟﺘﺤﺩﻴﺩ ﻗﻴﻡ ﺍﻟﻤﻌﺎﻤﻼﺕ ﺍﻷﻭﻟﻰ ﻤﻨﻬﺎ
ﻭﺭﺅﻴﺔ ﻤﺎ ﺇﺫﺍ ﻜﺎﻨﺕ ﺍﻷﺨﻁﺎﺀ residualsﻏﻴﺭ ﻤﺘﺭﺍﺒﻁﺔ ﺫﺍﺘﻴﹰﺎ ،ﺃﻱ ﺃﻥ ﻗﻴﻡ ﻤﻌﺎﻤﻼﺕ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ ﺘﺴﺎﻭﻱ ﺍﻟﺼﻔﺭ ﻓﻲ ﺤﺩﻭﺩ 95%ﻓﺘﺭﺓ ﺜﻘﺔ ﻭﻻ
ﺘﺤﺘﻭﻱ ﻫﺫﻩ ﺍﻟﻤﻌﺎﻤﻼﺕ ﻋﻠﻰ ﻨﻤﻁ ﻤﻌﻴﻥ ،ﻭﻋﻨﺩﻤﺎ ﺘﻜﻭﻥ ﺍﻟﻘﻴﻡ ﺍﻟﻤﻘﺩﺭﺓ ﻗﺭﻴﺒﺔ ﺠﺩﹰﺍ ﻤﻥ ﺍﻟﻘﻴﻡ ﺍﻟﺤﻘﻴﻘﻴﺔ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﻓﺈﻥ ﻤﺠﻤﻭﻉ ﻤﺭﺒﻌﺎﺕ ﺍﻷﺨﻁﺎﺀ ﺴﻭﻑ ﻴﻜﻭﻥ ﺍﻗل ﻤﺎ
ﻴﻤﻜﻥ )ﻁﺭﻴﻘﺔ ﺍﻟﻤﺭﺒﻌﺎﺕ ﺍﻟﺼﻐﺭﻯ( ،ﻭﻟﻥ ﻴﻜﻭﻥ ﻤﺘﻭﺴﻁ ﺍﻷﺨﻁﺎﺀ ﻤﺨﺘﻠﻑ ﻤﻌﻨﻭﻴﹰﺎ ﻋﻥ ﺍﻟﺼﻔﺭ ،ﻭﻜﺫﻟﻙ ﻴﻤﻜﻥ ﻓﺤﺹ ﻨﻤﺎﺫﺝ ﺒﺩﻴﻠﺔ ﻭﻤﻘﺎﺭﻨﺔ ﺍﻟﺘﻁﻭﺭ ﻓﻲ ﻫﺫﻩ ﺍﻟﻌﻭﺍﻤل ﻤﻊ
ﻤﺭﺍﻋﺎﺓ ﺃﻨﻪ ﻴﻔﻀل ﻋﺎﺩﺓ ﺍﺨﺘﻴﺎﺭ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺫﻱ ﻴﺤﺘﻭﻱ ﻋﻠﻰ ﺃﻗل ﻋﺩﺩ ﻤﻤﻜﻥ ﻤﻥ
ﺍﻟﻤﻌﺎﻤﻼﺕ ﻭﻴﻜﻭﻥ ﻤﻨﺎﺴﺒﹰﺎ ﻟﻠﺒﻴﺎﻨﺎﺕ ،ﻜﻤﺎ ﺃﻨﻪ ﻴﺠﺏ ﺃﻻ ﻴﻜﻭﻥ ﻫﻨﺎﻙ ﺍﺭﺘﺒﺎﻁﹰﺎ ﻤﻠﺤﻭﻅﹰﺎ
ﺒﻴﻥ ﺍﻟﻤﻌﺎﻟﻡ ﺍﻟﻤﺨﺘﻠﻔﺔ ﻟﻠﻨﻤﻭﺫﺝ ،ﻭﻜﺫﻟﻙ ﻻ ﻴﺠﺏ ﺃﻥ ﺘﺤﺘﻭﻱ ﺤﺩﻭﺩ ﺍﻟﺜﻘﺔ ﺒﺩﺍﺨﻠﻬﺎ ﻋﻠﻰ ﺍﻟﺼﻔﺭ ﻤﻤﺎ ﻴﻌﻨﻲ ﺃﻨﻪ ﻴﺠﺏ ﺃﻥ ﺘﻜﻭﻥ ﺍﻟﻤﻌﺎﻟﻡ ﻤﺨﺘﻠﻔﺔ ﻤﻌﻨﻭﻴﹰﺎ ﻋﻥ ﺍﻟﺼﻔﺭ ،ﻭﻓﻲ ﺍﻟﻨﻬﺎﻴﺔ
ﺇﺫﺍ ﺜﺒﺕ ﺃﻥ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺴﺎﻜﻥ ﻤﻨﺎﺴﺏ ﻟﻠﺴﻠﺴﻠﺔ ﻓﺈﻨﻪ ﻴﻤﻜﻥ ﺍﻻﻋﺘﻤﺎﺩ ﻋﻠﻴﻪ ﻓﻲ ﺍﻟﺘﻨﺒﺅ.
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
526
) : ARMA(2,1ﺇﺫﺍ ﺜﺒﺕ ﺃﻥ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺴﺎﺒﻕ ) ARMA(1,0ﻏﻴﺭ ﻤﻼﺌﻡ ﻟﺒﻴﺎﻨﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﻤﺤل ﺍﻟﺩﺭﺍﺴﺔ ﺒﻐﻴﺎﺏ ﺍﻟﺸﺭﻭﻁ ﺍﻟﻤﻁﻠﻭﺒﺔ ﻟﺫﻟﻙ ﺍﻟﻨﻤﻭﺫﺝ ﺒﻤﻌﺎﻤﻼﺕ ﺍﻷﺨﻁﺎﺀ ﻗﺭﻴﺒﺔ ﻤﻥ ﺍﻟﺼﻔﺭ ﻓﺈﻨﻪ ﻴﻤﻜﻥ ﺍﻟﺘﻘﺩﻡ ﺍﻵﻥ ﺨﻁﻭﺓ ﻓﻲ ﺍﻟﻁﺭﻴﻘﺔ ﺍﻟﻤﺤﺴﻨﺔ ﻹﺠﺭﺍﺀ
ﺘﺤﻠﻴل ﻟﻨﻤﺎﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﺍﻟﻤﺘﻜﺎﻤﻠﺔ ﻭﺫﻟﻙ ﺒﺘﻁﺒﻴﻕ ﻨﻤﻭﺫﺝ ) ARMA(2,1ﺍﻟﺫﻱ ﻴﺸﻤل ﻋﻠﻰ ﻤﻌﻠﻤﺘﻴﻥ ﺍﻨﺤﺩﺍﺭ ﺫﺍﺘﻲ ARﻭﻤﻌﻠﻤﺔ ﻭﺍﺤﺩﺓ ﻤﺘﻭﺴﻁﺎﺕ ﻤﺘﺤﺭﻜﺔ MAﺜﻡ
ﻓﺤﺹ ﻨﻤﻁ ﺍﻷﺨﻁﺎﺀ residualsﻜﻤﺎ ﺴﺒﻕ ﻓﻲ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺴﺎﺒﻕ.
ﺍﻟﺨﻁﻭﺍﺕ ﺍﻟﺘﺎﻟﻴﺔ :ﻁﺎﻟﻤﺎ ﺃﻨﻪ ﺒﻔﺤﺹ ﺸﺭﻭﻁ ﻨﻤﻁ ﺍﻷﺨﻁﺎﺀ residualsﻜﻤﺎ ﺴﺒﻕ ﺘﻭﻀﻴﺤﻬﺎ ﻓﻲ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﺴﺎﺒﻘﺔ ﻴﺘﺒﻴﻥ ﺃﻥ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺫﻱ ﻴﺘﻡ ﺍﺨﺘﻴﺎﺭﻩ ﻟﻠﺴﻠﺴﻠﺔ
ﺍﻟﺯﻤﻨﻴﺔ ﻟﻴﺱ ﻤﻼﺌﻤﹰﺎ ﺒﻌﺩ ﻓﺈﻥ ﺘﺤﻠﻴل ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﻻﺒﺩ ﻭﺃﻥ ﻴﺴﺘﻤﺭ ﺒﺎﺨﺘﻴﺎﺭ ﻨﻤﺎﺫﺝ
ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ) ARMA (n, n-1ﻭﺍﻟﺘﺩﺭﺝ ﻓﻲ ﺍﻟﻤﻌﻠﻤﺘﻴﻥ n, n-1ﻓﻲ ﺍﻟﻨﻤﻭﺫﺝ ﻭﺘﻁﺒﻴﻘﻪ ﻭﻤﻥ ﺜﻡ ﻓﺤﺹ ﺍﻷﺨﻁﺎﺀ ﻟﺤﻴﻥ ﺍﻟﺤﺼﻭل ﻋﻠﻰ ﻨﻤﻭﺫﺝ ﻤﻼﺌﻡ ﻟﻬﺫﻩ ﺍﻟﺴﻠﺴﻠﺔ،
ﻭﺃﺜﻨﺎﺀ ﺘﻁﺒﻴﻕ ﺘﻠﻙ ﺍﻟﻌﻤﻠﻴﺔ ﺇﺫﺍ ﺘﺒﻴﻥ ﺃﻥ ﻤﻌﺎﻤل ﺍﻻﻨﺤﺩﺍﺭ ﺍﻟﺫﺍﺘﻲ ) ϕﻓﺎﻱ (phiﻴﻘﺘﺭﺏ ﻤﻥ ﺍﻟﺼﻔﺭ ﻓﺈﻥ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺘﺎﻟﻲ ﺍﻟﺫﻱ ﻻ ﺒﺩ ﻤﻥ ﺍﺨﺘﻴﺎﺭﻩ ﻭﺍﻟﺘﺤﻘﻕ ﻤﻨﻪ ﻜﻤﺎ ﺴﺒﻕ ﻴﻜﻭﻥ
) ، ARMA (n-1,n-1ﻭﺒﺎﻟﻤﺜل ﺇﺫﺍ ﺘﺒﻴﻥ ﺃﻥ ﻤﻌﺎﻤل ﺍﻟﻤﺘﻭﺴﻁﺎﺕ ﺍﻟﻤﺘﺤﺭﻜﺔ ) θﺜﻴﺘﺎ
(thetaﻴﻘﺘﺭﺏ ﻤﻥ ﺍﻟﺼﻔﺭ ﻓﺈﻥ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺘﺎﻟﻲ ﺍﻟﺫﻱ ﻻ ﺒﺩ ﻤﻥ ﺍﺨﺘﻴﺎﺭﻩ ﻭﺍﻟﺘﺤﻘﻕ ﻤﻨﻪ
ﻴﺠﺏ ﺃﻥ ﻴﻜﻭﻥ ) ، ARMA (n,n-2ﻭﺃﺜﻨﺎﺀ ﺘﻔﺤﺹ ﻨﻤﺎﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﺍﻟﻤﺘﺘﺎﻟﻴﺔ
ﻭﻓﻲ ﻟﺤﻅﺔ ﻤﻌﻴﻨﺔ ﻗﺩ ﺘﺘﻼﺸﻰ ﻤﻌﺎﻤﻼﺕ ﺃﻱ ﻤﻥ ﺍﻻﻨﺤﺩﺍﺭ ﺍﻟﺫﺍﺘﻲ ﺃﻭ ﺍﻟﻤﺘﻭﺴﻁﺎﺕ
ﺍﻟﻤﺘﺤﺭﻜﺔ ،ﻓﻲ ﻤﺜل ﻫﺫﻩ ﺍﻟﺤﺎﻟﺔ ﺇﻥ ﻭﺼﻠﻨﺎ ﺇﻟﻴﻬﺎ ﻓﺈﻨﻪ ﻴﺠﺏ ﺍﻻﺴﺘﻤﺭﺍﺭ ﻓﻲ ﺍﺨﺘﻴﺎﺭ
ﻭﻓﺤﺹ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﺴﺎﻜﻨﺔ ﺍﻟﻤﺘﺘﺎﻟﻴﺔ ﺒﺎﻟﻤﻌﺎﻟﻡ ﺍﻟﻤﺘﺒﻘﻴﺔ ﻭﺍﻟﺘﺤﻘﻕ ﻤﻥ ﻤﻼﺀﻤﺘﻬﺎ ﻟﺤﻴﻥ ﺃﻥ ﺘﻘﺘﺭﺏ ﻤﻌﺎﻤﻼﺕ ﺍﻷﺨﻁﺎﺀ ﻤﻌﻨﻭﻴﹰﺎ ﻤﻥ ﺍﻟﺼﻔﺭ ﻟﺘﻘﻊ ﺒﻴﻥ ﺤﺩﻱ 95%ﻓﺘﺭﺓ ﺍﻟﺜﻘﺔ ﻭﺘﺤﻘﻕ
ﺃﻴﻀﹰﺎ ﺍﻟﺸﺭﻭﻁ ﺍﻟﻤﻁﻠﻭﺒﺔ ﺍﻟﺴﺎﺒﻘﺔ ﻟﻜﻲ ﻨﺼل ﺇﻟﻰ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﻤﻨﺎﺴﺏ.
ﻭﻴﻤﻜﻥ ﺘﻠﺨﻴﺹ ﺍﻟﺨﻁﻭﺍﺕ ﺍﻟﺴﺎﺒﻘﺔ ﺒﺎﻟﺭﺴﻡ ﺍﻟﺘﻭﻀﻴﺤﻲ ﻓﻲ ﺸﻜل 2-15ﺃﺩﻨﺎﻩ.
( ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ15)
527
ﺠﻴﻨﻜﻨﺯ ﺍﻟﻤﻼﺌﻡ- ﺨﻁﻭﺍﺕ ﺍﺨﺘﻴﺎﺭ ﻨﻤﻭﺫﺝ ﺒﻭﻜﺱ: 2-15 ﺸﻜل ARMA (1,0)?
ARMA (2,0)?
ARMA (2,1)?
ARMA (1,1)?
ARMA (0,1)?
ARMA (3,0)?
ARMA (3,2)?
ARMA (2,2)? ARMA (1,2)? ARMA (0,2)?
ARMA (n,0)?
ARMA (n,n-1)?
ARMA (n-1,n-1)?
ARMA (n-2,n-1)?
ARMA (0,n-1)?
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
528
ﺍﻟﺘﺤﻠﻴل ﺍﻟﻤﻭﺴﻤﻲ :ﺒﻨﻔﺱ ﺍﻟﻁﺭﻴﻘﺔ ﺍﻟﺴﺎﺒﻘﺔ ﻓﻲ ﺘﻁﻭﺭ ﻋﻤﻠﻴﺔ ﺍﺨﺘﻴﺎﺭ ﻨﻤﻭﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﺍﻟﻤﻼﺌﻡ ﻟﻠﺒﻴﺎﻨﺎﺕ ﻭﻤﺤﺎﺫﺍﺓ ﻟﺘﻠﻙ ﺍﻟﺨﻁﻭﺍﺕ ﻴﻤﻜﻥ ﺇﻀﺎﻓﺔ ﺃﻭ ﺤﺫﻑ ﺃﻱ
ﻤﻥ ﻤﻌﺎﻟﻡ ﺍﻻﻨﺤﺩﺍﺭ ﺍﻟﺫﺍﺘﻲ ﺍﻟﻤﻭﺴﻤﻲ SARﺃﻭ ﻤﻌﺎﻟﻡ ﺍﻟﻤﺘﻭﺴﻁﺎﺕ ﺍﻟﻤﺘﺤﺭﻜﺔ ﺍﻟﻤﻭﺴﻤﻴﺔ
SMAﺍﺴﺘﺠﺎﺒﺔ ﻟﻭﺠﻭﺩ ﺃﻭ ﻏﻴﺎﺏ ﻨﻤﻁ ﻤﻭﺴﻤﻲ ﺃﻭ ﺩﻭﺭﻱ ﻓﻲ ﻗﻴﻡ ﺍﻷﺨﻁﺎﺀ ﺃﻭ ﺃﻥ ﻗﻴﻡ ﺍﻟﻤﻌﺎﻟﻡ ﺘﺘﻘﺎﺭﺏ ﻤﻥ ﺍﻟﺼﻔﺭ.
ﻤﻼﺌﻤﺔ ﺍﻟﻨﻤﻭﺫﺝ :ﺃﺜﻨﺎﺀ ﻤﺭﺤﻠﺔ ﻤﺭﺍﺠﻌﺔ ﻨﺘﺎﺌﺞ ﺘﺤﻠﻴل ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﻟﻠﺴﻠﺴﻠﺔ
ﺍﻟﺯﻤﻨﻴﺔ ﻻﺒﺩ ﻤﻥ ﺃﺨﺫ ﺍﻟﺤﻴﻁﺔ ﻭﺍﻟﺤﺫﺭ ﻟﻠﺘﺄﻜﺩ ﻤﻥ ﺃﻥ ﻤﻌﺎﻟﻡ ﺍﻟﻨﻤﻭﺫﺝ ﻏﻴﺭ ﻤﺘﺭﺍﺒﻁﺔ ،ﻜﻤﺎ ﺃﻨﻪ ﻴﺠﺏ ﻤﻌﺎﻴﺭﺓ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﻤﻌﻨﻭﻴﺔ ﻭﺍﻟﺒﺩﻴﻠﺔ ﻤﻘﺎﺒل ﻫﺫﻩ ﺍﻟﺸﺭﻭﻁ ﻭﻜﺫﻟﻙ ﻤﻘﺎﺒل ﻗﻴﻤﺔ
ﻤﺭﺒﻊ ﻤﻌﺎﻤل ﺍﻻﺭﺘﺒﺎﻁ ﺍﻟﻌﺎﻡ R2ﻭﺍﻟﺨﻁﺄ ﺍﻟﻤﻌﻴﺎﺭﻱ ﻭﺘﻘﺎﺭﺏ ﻗﻴﻡ ﺍﻷﺨﻁﺎﺀ ﻤﻥ ﺍﻟﺼﻔﺭ
ﺒﺩﺭﺠﺔ ﺜﻘﺔ ﻤﻨﺎﺴﺒﺔ.
ﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ ﺒﺎﺴﺘﺨﺩﺍﻡ ﻨﻤﻭﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﺍﻟﺫﻱ ﺘﻡ ﺘﻁﻭﻴﺭﻩ :ﻟﻘﺩ ﺘﻡ
ﺍﻟﺒﺤﺙ ﻋﻥ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺫﻱ ﻴﻤﻜﻥ ﺃﻥ ﻴﺼﻑ ﺍﻟﺘﻐﻴﺭ ﻓﻲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﻤﻥ ﺃﺠل ﺍﺴﺘﺨﺩﺍﻤﻬﺎ ﻓﻲ ﺍﻟﺘﻨﺒﺅ ﺒﻘﻴﻡ ﻤﺴﺘﻘﺒﻠﻴﺔ ﻗﺼﻴﺭﺓ ﻭﻤﺘﻭﺴﻁﺔ ﺍﻷﺠل ،ﻭﻻ ﻴﺠﺏ ﺍﺴﺘﺨﺩﺍﻡ ﻤﺜل
ﻫﺫﺍ ﺍﻟﻨﻤﻭﺫﺝ ﻓﻲ ﺍﻟﺘﻨﺒﺅ ﺒﻘﻴﻡ ﻤﺴﺘﻘﺒﻠﻴﺔ ﻁﻭﻴﻠﺔ ﺍﻷﺠل ،ﻭﻴﺤﺴﻥ ﺩﺍﺌﻤﹰﺎ ﺘﻌﺩﻴل ﺍﻟﺒﻴﺎﻨﺎﺕ ﻜﻠﻤﺎ ﺘﻭﻓﺭﺕ ﻤﻌﻠﻭﻤﺎﺕ ﺠﺩﻴﺩﺓ ﻋﻥ ﻓﺘﺭﺍﺕ ﺯﻤﻨﻴﺔ ﺠﺩﻴﺩﺓ ﻟﻜﻲ ﻴﺘﻡ ﺘﺨﻔﻴﺽ ﻋﺩﺩ ﺍﻟﻔﺘﺭﺍﺕ
ﺍﻟﺯﻤﻨﻴﺔ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﺍﻟﻤﻁﻠﻭﺏ ﺍﻟﺘﻨﺒﺅ ﺒﻘﻴﻡ ﺍﻟﺴﻠﺴﻠﺔ ﺒﻬﺎ.
ﻤﺭﺍﻗﺒﺔ ﺩﻗﺔ ﺍﻟﺘﻨﺒﺅ ﻋﻤﻠﻴ ﹰﺎ :ﻜﻠﻤﺎ ﻤﺭ ﺍﻟﻭﻗﺕ ﻓﺈﻨﻪ ﻴﺠﺏ ﻤﺭﺍﻗﺒﺔ ﻤﺩﻯ ﺩﻗﺔ ﻨﻤﻭﺫﺝ
ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﺍﻟﺫﻱ ﺘﻡ ﺘﻁﻭﻴﺭﻩ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﻓﻲ ﺍﻟﺘﻨﺒﺅ ﺒﻘﻴﻡ ﺍﻟﻅﺎﻫﺭﺓ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ
ﺍﻟﺘﻲ ﺃﺼﺒﺤﺕ ﺤﺎﻟﻴﺔ ﻭﻤﺸﺎﻫﺩﺓ ،ﻓﻜﻠﻤﺎ ﻜﺎﻨﺕ ﺍﻟﻔﺘﺭﺓ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﺒﻌﻴﺩﺓ ﻜﻠﻤﺎ ﻜﺎﻥ ﻋﺎﻤل
ﺍﻟﺨﻁﺄ ﻓﻲ ﺍﻟﺘﻨﺒﺅ ﺃﻜﺒﺭ ،ﻟﺫﻟﻙ ﻻﺒﺩ ﻤﻥ ﻤﺭﺍﻗﺒﺔ ﺘﻁﻭﺭ ﺍﻷﺨﻁﺎﺀ ﻓﻲ ﺍﻟﺘﻨﺒﺅ ﻭﺍﻟﺨﻁﺄ
ﺍﻟﻤﻌﻴﺎﺭﻱ ﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﻟﻸﺨﻁﺎﺀ ،ﻭﺇﺫﺍ ﺘﺒﻴﻥ ﺃﻥ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺒﺩﺃﺕ ﺘﺄﺨﺫ ﻨﻤﻁﹰﺎ
ﺠﺩﻴﺩﹰﺍ ﻤﻊ ﻤﺭﻭﺭ ﺍﻟﻭﻗﺕ ﻓﻼﺒﺩ ﻫﻨﺎ ﻤﻥ ﺇﻋﺎﺩﺓ ﺘﻘﺩﻴﺭ ﻤﻌﺎﻟﻡ ﺍﻟﻨﻤﻭﺫﺝ ﻤﻥ ﺠﺩﻴﺩ.
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
529
.5 .15اﺧﺘﻴﺎر ﻧﻤﻮذج اﻟﺴﻼﺳﻞ اﻟﺰﻣﻨﻴﺔ ﺑﺎﺳﺘﺨﺪام ﻧﻈﺎم : SPSS Model Identification Using SPSS: ﻭﻟﺘﻭﻀﻴﺢ ﺍﻷﺴﺱ ﺍﻟﺴﺎﺒﻘﺔ ﺴﻨﺄﺨﺫ ﺍﻟﻤﺜﺎل ﺍﻟﺘﺎﻟﻲ ﻭﺍﻟﺫﻱ ﻴﻬﺩﻑ ﺇﻟﻰ ﺘﻭﻀﻴﺢ ﺍﻟﺒﻨﻭﺩ ﺍﻟﺴﺎﺒﻕ ﺫﻜﺭﻫﺎ ﺒﺎﺴﺘﺨﺩﺍﻡ ﺒﻴﺎﻨﺎﺕ ﻟﺴﻠﺴﻠﺔ ﺯﻤﻨﻴﺔ ﺘﺘﻜﻭﻥ ﻤﻥ 48ﻗﻴﻤﺔ ﻭﺘﻤﺜل
ﺴﻠﺴﻠﺔ ﺍﻷﺭﻗﺎﻡ ﺍﻟﻘﻴﺎﺴﻴﺔ ﺍﻟﺸﻬﺭﻴﺔ ﻷﺴﻌﺎﺭ ﺍﻟﻤﺴﺘﻬﻠﻙ ﻓﻲ ﻓﻠﺴﻁﻴﻥ ﻓﻲ ﺍﻷﺭﺒﻊ ﺴﻨﻭﺍﺕ 2000-1997ﻤﺒﻴﻨﺔ ﻓﻲ ﺍﻟﺠﺩﻭل ﻓﻲ ﺸﻜل 3-15ﺍﻟﺘﺎﻟﻲ.
ﺸﻜل : 3-15ﺴﻠﺴﻠﺔ ﺯﻤﻨﻴﺔ ﺘﻤﺜل ﺍﻷﺭﻗﺎﻡ ﺍﻟﻘﻴﺎﺴﻴﺔ ﺍﻟﺸﻬﺭﻴﺔ ﻟﻠﻤﺴﺘﻬﻠﻙ Monthly Consumer Price Indexﻓﻲ 48ﺸﻬﺭﹰﺍ ﻟﻸﻋﻭﺍﻡ 2000-1997ﻓﻲ ﻓﻠﺴﻁﻴﻥ. ) ( 1996 =100 ﻋﺎﻡ 2000 ﻋﺎﻡ 1999 ﻋﺎﻡ 1997ﻋﺎﻡ 1998 ﺍﻟﺸﻬﺭ ﻴﻨﺎﻴﺭ
104.74
111.86
121.76
123.73
ﻓﺒﺭﺍﻴﺭ
106.35
111.46
120.00
124.07
ﻤﺎﺭﺱ
106.25
110.62
119.79
123.79
ﺇﺒﺭﻴل
107.61
110.31
118.33
122.96
ﻤﺎﻴﻭ
106.21
111.28
118.27
132.27
ﻴﻭﻨﻴﻭ
106.77
111.19
119.77
123.30
ﻴﻭﻟﻴﻭ
107.98
111.75
119.17
123.09
ﺃﻏﺴﻁﺱ
108.54
112.39
118.56
121.95
ﺴﺒﺘﻤﺒﺭ
108.96
114.62
118.78
122.86
ﺃﻜﺘﻭﺒﺭ
108.97
117.89
120.81
122.95
ﻨﻭﻓﻤﺒﺭ
109.19
119.60
121.35
123.19
ﺩﻴﺴﻤﺒﺭ
109.43
120.57
123.55
126.25
ﻭﻟﺘﻭﻀﻴﺢ ﻜﻴﻔﻴﺔ ﺇﻴﺠﺎﺩ ﻨﻤﻭﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﺍﻟﻤﻼﺌﻡ ﻟﻬﺫﻩ ﺍﻟﺴﻠﺴﻠﺔ ﻻﺒﺩ ﻓﻲ ﺍﻟﺒﺩﺍﻴﺔ ﻤﻥ ﺘﻤﺜﻴل ﻫﺫﻩ ﺍﻟﺴﻠﺴﻠﺔ ﺒﻴﺎﻨﻴﹰﺎ ﺒﺎﺴﺘﺨﺩﺍﻡ ﺸﻜل ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﻋﻥ ﻁﺭﻴﻕ ﺃﻤﺭ
ﺭﺴﻡ ﺍﻟﺴﻠﺴﻠﺔ Sequenceﻓﻲ ﻗﺎﺌﻤﺔ ﺍﻟﺭﺴﻭﻤﺎﺕ ﺍﻟﺒﻴﺎﻨﻴﺔ Graphsﻓﻲ ﻨﻅﺎﻡ ، SPSS ﻭﻴﺒﻴﻥ ﺍﻟﺸﻜل 4-15ﺍﻟﺘﺎﻟﻲ ﺭﺴﻤﹰﺎ ﻟﻬﺫﻩ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ.
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
530
ﺸﻜل : 4-15ﺴﻠﺴﻠﺔ ﺯﻤﻨﻴﺔ ﺘﻤﺜل ﺍﻷﺭﻗﺎﻡ ﺍﻟﻘﻴﺎﺴﻴﺔ ﺍﻟﺸﻬﺭﻴﺔ ﻟﻠﻤﺴﺘﻬﻠﻙ Monthly Consumer Price Indexﻓﻲ 48ﺸﻬﺭﹰﺍ ﻟﻸﻋﻭﺍﻡ 2000-1997ﻓﻲ ﻓﻠﺴﻁﻴﻥ. Monthly Consumer Price Index in Palestine 1997-2000 1996 = 100 Source: Statistical Abstract of Palestine, No. (2), 2001 PRICIND
140
130
120
110
100 2000
1999
1998
1997
Month and Year
ﻭﻏﻨﻲ ﻋﻥ ﺍﻟﺒﻴﺎﻥ ﺃﻥ ﻫﺫﺍ ﺍﻟﺭﺴﻡ ﻭﺠﻤﻴﻊ ﺍﻟﺒﻨﻭﺩ ﻓﻲ ﺍﻟﺨﻁﻭﺍﺕ ﺍﻟﺘﺎﻟﻴﺔ ﻗﺩ ﺘﻤﺕ
ﺒﺎﺴﺘﺨﺩﺍﻡ ﻨﻅﺎﻡ SPSSﻋﻥ ﻁﺭﻴﻕ ﺍﻷﻭﺍﻤﺭ ﺍﻟﺘﻲ ﺘﻡ ﺘﻭﻀﻴﺤﻬﺎ ﻓﻲ ﺍﻟﻔﺼﻭل ﺍﻟﺴﺎﺒﻘﺔ
ﺒﺎﺴﺘﺜﻨﺎﺀ ﺒﻌﺽ ﺍﻷﻭﺍﻤﺭ ﺍﻟﺠﺩﻴﺩﺓ ﺍﻟﺘﻲ ﺴﻴﺘﻡ ﺍﻟﺘﻌﺭﺽ ﺇﻟﻴﻬﺎ ﻓﻴﻤﺎ ﻴﻠﻲ.
ﺇﻥ ﺃﺤﺩ ﺍﻷﺩﻭﺍﺕ ﺍﻟﻬﺎﻤﺔ ﻓﻲ ﺍﺴﺘﻜﺸﺎﻑ ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﻓﺤﺹ ﺸﺭﻭﻁ ﺍﻟﺜﺒﺎﺕ
ﻭﺍﺴﺘﻘﺭﺍﺭ ﺍﻟﺘﺒﺎﻴﻥ ﻫﻭ ﺘﻤﺜﻴل ﻫﺫﻩ ﺍﻟﺒﻴﺎﻨﺎﺕ ﺒﺎﺴﺘﺨﺩﺍﻡ ﺃﺸﻜﺎل ﺍﻟﺼﻨﺎﺩﻴﻕ ، Box Plots ﻭﻫﺫﺍ ﻴﻤﻜﻥ ﺃﻥ ﻴﺘﻡ ﺒﻌﺩ ﺘﻘﺴﻴﻡ ﺍﻟﺴﻠﺴﻠﺔ ﺇﻟﻰ ﻤﺠﻤﻭﻋﺎﺕ ﻤﻥ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺘﺘﺎﺒﻌﺔ ) 12ﻗﻴﻤﺔ ﻫﻨﺎ( ﻭﺇﻋﻁﺎﺀ ﺩﻟﻴل ﻟﻜل ﻤﺠﻤﻭﻋﺔ)ﻓﺌﺔ(.
ﻭﻴﻭﻀﺢ ﺍﻟﺸﻜل 5-15ﺍﻟﺘﺎﻟﻲ ﺃﺸﻜﺎل ﺍﻟﺼﻨﺎﺩﻴﻕ Box plotﻟﻬﺫﻩ ﺍﻟﺴﻠﺴﻠﺔ،
ﻓﻴﺘﻀﺢ ﻤﻨﻪ ﻭﻤﻥ ﺍﻟﺸﻜل ﺍﻟﺴﺎﺒﻕ ﺃﻨﻪ ﻻ ﻴﻭﺠﺩ ﺤﻘﻴﻘﺔ ﺍﺘﺠﺎﻩ ﻋﺎﻡ ﺨﻁﻲ Linear Trend
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
531
ﻓﻲ ﻫﺫﻩ ﺍﻟﺴﻠﺴﺔ ﺒل ﻴﻤﻜﻥ ﺃﻥ ﻨﺼﻑ ﺍﻟﺘﻐﻴﺭ ﻓﻲ ﻫﺫﻩ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺭﺘﺒﻁ ﺒﺎﻟﺯﻤﻥ ﺒﺄﻨﻪ ﺍﺘﺠﺎﻩ ﻋﺎﻡ ﻴﺄﺨﺫ ﺸﻜل ﻤﻨﺤﻨﻰ ﻤﻥ ﺍﻟﺩﺭﺠﺔ ﺍﻟﺜﺎﻨﻴﺔ Quadratic Curveﺃﻭ ﺒﺸﻜل ﺃﺩﻕ ﻴﻤﻜﻥ ﺃﻥ ﻴﻭﺼﻑ ﺒﺄﻨﻪ ﺘﺤﻭل ﻤﻔﺎﺠﺊ a step changeﻓﻲ ﻤﻨﺘﺼﻑ ﺍﻟﺴﻠﺴﻠﺔ ،ﻟﺫﻟﻙ ﺴﻭﻑ ﻨﻘﻭﻡ ﺒﺘﺠﺎﻫل ﻭﺠﻭﺩﻩ ﻓﻲ ﻫﺫﻩ ﺍﻟﻤﺭﺤﻠﺔ .
ﺸﻜل : 5-15ﺃﺸﻜﺎل ﺍﻟﺼﻨﺎﺩﻴﻕ Box Plotsﻟﺴﻠﺴﻠﺔ ﺍﻷﺭﻗﺎﻡ ﺍﻟﻘﻴﺎﺴﻴﺔ ﺍﻟﺸﻬﺭﻴﺔ ﻟﻠﻤﺴﺘﻬﻠﻙ Consumer Price Indexﻟﻸﻋﻭﺍﻡ 2000-1997ﻓﻲ ﻓﻠﺴﻁﻴﻥ. Box plot for monthly consumer price index in Palestine 1997-2000 1996 = 100
41
130 48
120
Monthly Consumer Price Index
140
110
100 12
12
12
12
2000
1999
1998
1997
=N
Year
ﻭﻻﺴﺘﻜﺸﺎﻑ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﻤﻨﺎﺴﺏ ﻟﻬﺫﻩ ﺍﻟﺴﻠﺴﻠﺔ ﻻ ﺒﺩ ﻤﻥ ﺭﺴﻡ ﺩﺍﻟﺘﻲ ﺍﻟﺘﺭﺍﺒﻁ
ﺍﻟﺫﺍﺘﻲ ACFﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ PACFﻜﻤﺎ ﻓﻲ ﺸﻜل 6-15ﺃﺩﻨﺎﻩ.
ﻴﺘﻀﺢ ﻤﻥ ﻨﻤﻁ ﺩﺍﻟﺘﻲ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ACFﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ PACF
ﻓﻲ ﺍﻟﺸﻜل ﺍﻟﺘﺎﻟﻲ ﺃﻥ ﻨﻤﻭﺫﺝ ﺍﻻﻨﺤﺩﺍﺭ ﺍﻟﺫﺍﺘﻲ ﻤﻥ ﺍﻟﺩﺭﺠﺔ ﺍﻷﻭﻟﻰ ) AR(1ﻫﻭ ﺍﻟﻨﻤﻭﺫﺝ
ﺍﻷﻜﺜﺭ ﻤﻼﺌﻤﺔ ﻟﻬﺫﻩ ﺍﻟﺒﻴﺎﻨﺎﺕ ﺤﻴﺙ ﺃﻥ ﺩﺍﻟﺔ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ACFﺘﺘﻼﺸﻰ ﺘﺩﺭﻴﺠﻴﹰﺎ
ﺒﻴﻨﻤﺎ ﺩﺍﻟﺔ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ PACFﺘﺘﻭﻗﻑ ﺒﻌﺩ ﺍﻟﺩﺭﺠﺔ ﺍﻷﻭﻟﻰ.
( ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ15)
532
. ﻟﻠﺴﻠﺴﻠﺔPACF ﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲACF ﺩﺍﻟﺘﻲ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ: 6-15 ﺸﻜل ACF
Monthly Consumer Price Index 1.0
.5
0.0
-.5
Confidence Limits
Coefficient
-1.0 1
3 2
5 4
7 6
9 8
11 10
13 12
15 14
16
Lag Number
Partial ACF
Monthly Consumer Price Index 1.0
.5
0.0
-.5
Confidence Limits
Coefficient
-1.0 1
3 2
5 4
7 6
9 8
11 10
13 12
15 14
16
Lag Number
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
533
ﻭﻴﺠﺩﺭ ﺒﺎﻟﺫﻜﺭ ﺃﻥ ﺒﻴﺎﻨﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺘﺩﺨل ﻋﺎﺩﺓ ﻓﻲ ﻨﻅﺎﻡ SPSSﺩﺍﺌﻤﹰﺎ ﺒﺎﺴﻡ ﻤﺘﻐﻴﺭ ﻭﺍﺤﺩ ﻓﻲ ﻋﻤﻭﺩ ﻭﺍﺤﺩ ﻓﻘﻁ ﻟﺫﺍ ﻻ ﺒﺩ ﻤﻥ ﺇﺩﺨﺎل ﺒﻴﺎﻨﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﻓﻲ ﺍﻟﺸﻜل 3-15ﺍﻟﺴﺎﺒﻕ ﻓﻲ ﻋﻤﻭﺩ ﻭﺍﺤﺩ ﺤﺴﺏ ﺍﻟﺘﺩﺭﺝ ﺍﻟﺯﻤﻨﻲ ،ﻭﻴﻤﻜﻥ ﺍﻟﺤﺼﻭل ﻋﻠﻰ ﺭﺴﻡ
ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ )ﺸﻜل (5-15ﻤﻥ ﺨﻼل ﺃﻤﺭ ﺭﺴﻡ ﺍﻟﺴﻠﺴﻠﺔ Sequenceﻓﻲ ﻗﺎﺌﻤﺔ ﺍﻟﺭﺴﻭﻤﺎﺕ ﺍﻟﺒﻴﺎﻨﻴﺔ Graphsﻓﻲ ﺍﻟﻘﺎﺌﻤﺔ ﺍﻟﺭﺌﻴﺴﻴﺔ ﻟﻠﻨﻅﺎﻡ ،ﺃﻱ : Sequence .
Graphs ----Æ
ﻭﺫﻟﻙ ﻜﻤﺎ ﻓﻲ ﺸﻜل 7-15ﻟﺘﻔﺘﺢ ﻨﺎﻓﺫﺓ ﺭﺴﻡ ﺍﻟﺴﻠﺴﻠﺔ Sequenceﻓﻴﺘﻡ ﺒﻬﺎ
ﺇﺩﺨﺎل ﺍﺴﻡ ﺍﻟﻤﺘﻐﻴﺭ ﺍﻟﺫﻱ ﻴﺤﺘﻭﻱ ﻋﻠﻰ ﺒﻴﺎﻨﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ،ﻭﻴﻤﻜﻥ ﺃﻴﻀﹰﺎ ﺇﺩﺨﺎل ﻤﺘﻐﻴﺭ ﺁﺨﺭ ﻴﺤﺘﻭﻱ ﻋﻠﻰ ﺍﻟﻭﺤﺩﺍﺕ ﺍﻟﺯﻤﻨﻴﺔ ﻟﺘﻌﺭﻴﻑ ﺍﻟﻤﺤﻭﺭ ﺍﻷﻓﻘﻲ ﻓﻲ ﺍﻟﺸﻜل ﺇﻻ ﺃﻥ
ﻫﺫﺍ ﻟﻴﺱ ﻀﺭﻭﺭﻱ ﺇﺫ ﻴﻘﻭﻡ ﺍﻟﻨﻅﺎﻡ ﺒﺘﻤﺜﻴل ﺍﻟﻘﻴﻡ ﺒﻴﺎﻨﻴﹰﺎ ﺒﺎﻟﺘﺴﻠﺴل ﺍﻟﺯﻤﻨﻲ ﺍﻟﻁﺒﻴﻌﻲ .
ﻜﻤﺎ ﺃﻨﻪ ﻴﻤﻜﻥ ﺃﻴﻀﹰﺎ ﺍﻟﺤﺼﻭل ﻋﻠﻰ ﺩﺍﻟﺘﻲ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ACFﻭﺍﻟﺘﺭﺍﺒﻁ
ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ PACFﻟﻬﺫﻩ ﺍﻟﺴﻠﺴﻠﺔ ﺒﺎﺴﺘﺨﺩﺍﻡ ﻨﻅﺎﻡ SPSSﻋﻥ ﻁﺭﻴﻕ ﺃﻤﺭ ﺭﺴﻡ
ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ،Time Series Plotﻓﻴﻤﻜﻥ ﺍﻟﺤﺼﻭل ﻋﻠﻰ ﺍﻟﺭﺴﻤﻴﻥ ﻓﻲ ﺍﻟﺸﻜل 6-15ﺒﺎﺨﺘﻴﺎﺭ ﻗﺎﺌﻤﺔ ﺍﻟﺭﺴﻭﻤﺎﺕ ﺍﻟﺒﻴﺎﻨﻴﺔ Graphsﻓﻲ ﺍﻟﻘﺎﺌﻤﺔ ﺍﻟﺭﺌﻴﺴﻴﺔ ﻟﻠﻨﻅﺎﻡ ﻭﻤﻨﻬﺎ
ﻴﺘﻡ ﺍﺨﺘﻴﺎﺭ ﺃﻭﺍﻤﺭ ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ Time Seriesﻭﻤﻥ ﺜﻡ ﻨﺨﺘﺎﺭ ﺃﻤﺭ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ
) Autocorrelationsﻜﻤﺎ ﻓﻲ ﺸﻜل ،(7-15ﺃﻱ :
Time Series --Æ Autocorrelations
Graphs --Æ
ﻟﺘﻔﺘﺢ ﻨﺎﻓﺫﺓ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ Autocorrelationsﻓﻴﺘﻡ ﺒﻬﺎ ﺇﺩﺨﺎل ﺍﺴﻡ ﺍﻟﻤﺘﻐﻴﺭ ﺍﻟﺫﻱ
ﻴﺤﺘﻭﻱ ﻋﻠﻰ ﺒﻴﺎﻨﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﻓﻘﻁ.
ﻜﺫﻟﻙ ﻴﻤﻜﻥ ﺘﻭﻓﻴﻕ ﻨﻤﻭﺫﺝ ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﺍﻟﻤﻘﺘﺭﺡ ) AR(1ﺃﻭ ﺃﻱ ﻤﻥ ﻨﻤﺎﺫﺝ
ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﻟﻠﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻋﻥ ﻁﺭﻴﻕ ﻗﺎﺌﻤﺔ ﺍﻟﺘﺤﻠﻴل ﺍﻹﺤﺼﺎﺌﻲ ) Analyzeﺃﻭ
Statisticsﻓﻲ ﺇﺼﺩﺍﺭ (8.0ﻭﻤﻨﻬﺎ ﺍﺨﺘﻴﺎﺭ ﺃﻭﺍﻤﺭ ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ Time Series
ﻭﻤﻥ ﺜﻡ ﺍﻷﻤﺭ ) ARIMAﺸﻜل (8-15ﺍﻟﺘﺎﻟﻲ .
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
534
ﺸﻜل : 7-15ﺍﻟﺤﺼﻭل ﻋﻠﻰ ﺩﺍﻟﺘﻲ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ACFﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ PACFﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﻤﻥ ﻗﺎﺌﻤﺔ ﺍﻟﺭﺴﻭﻤﺎﺕ ﺍﻟﺒﻴﺎﻨﻴﺔ .Ghraphs
ﺸﻜل : 8-15ﺘﺤﻠﻴل ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺒﺎﺴﺘﺨﺩﺍﻡ ﻨﻤﺎﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ .ARIMA
ﻭﺒﺎﺨﺘﻴﺎﺭ ﺃﻤﺭ ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ Time Seriesﻭﻤﻥ ﺜﻡ ﺍﻷﻤﺭ ) ARIMAﻜﻤﺎ
ﺴﺒﻕ ﺘﻭﻀﻴﺤﻪ( ﺴﻭﻑ ﺘﻔﺘﺢ ﻨﺎﻓﺫﺓ ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ARIMAﺍﻟﻤﻭﻀﺤﺔ ﻓﻲ
ﺍﻟﺸﻜل 9-15ﺃﺩﻨﺎﻩ ،ﻓﻴﺘﻡ ﺇﻋﻁﺎﺀ ﺍﺴﻡ ﺍﻟﻤﺘﻐﻴﺭ ﺍﻟﺫﻱ ﻴﺤﺘﻭﻱ ﻋﻠﻰ ﻗﻴﻡ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﻓﻘﻁ ﻓﻲ ﻤﺭﺒﻊ ﺃﺴﻤﺎﺀ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺘﺎﺒﻌﺔ ، Dependentﺜﻡ ﻴﺘﻡ ﺘﺤﺩﻴﺩ ﻗﻴﻡ p, d, q
ﻭﻫﻲ ﺭﺘﺒﺔ ﻨﻤﻭﺫﺝ ) ARIMA (p, d, qﻭﻫﻲ ﻫﻨﺎ 1ﻭ 0ﻭ 0ﻋﻠﻰ ﺍﻟﺘﺭﺘﻴﺏ.
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
535
ﺸﻜل : 9-15ﻨﺎﻓﺫﺓ ﻨﻤﺎﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ARIMAﻟﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ.
ﻱ ﻤﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻨﺎﺘﺠﺔ ﺇﻀﺎﻓﺔ ﺇﻟﻰ ﺫﻟﻙ ﻴﻤﻜﻥ ﻓﻲ ﺍﻟﻨﺎﻓﺫﺓ ﺍﻟﺴﺎﺒﻘﺔ ﺘﺤﺩﻴﺩ ﺃ ٍ
ﻟﻴﻘﻭﻡ ﺍﻟﻨﻅﺎﻡ ﺒﺘﺨﺯﻴﻨﻬﺎ ﻭﻋﺭﻀﻬﺎ ﻓﻲ ﻤﺤﺭﺭ ﺍﻟﺒﻴﺎﻨﺎﺕ ﻹﺠﺭﺍﺀ ﺘﺤﻠﻴل ﺇﻀﺎﻓﻲ ﻋﻠﻴﻬﺎ
ﻭﺫﻟﻙ ﻤﻥ ﺨﻼل ﺍﻟﻀﻐﻁ ﻋﻠﻰ ﻤﻔﺘﺎﺡ ﺤﻭﺍﺭ ﺍﻟﺘﺨﺯﻴﻥ Saveﺍﻟﺫﻱ ﺴﻴﻘﻭﻡ ﺒﻔﺘﺢ ﻨﺎﻓﺫﺓ ﺘﺨﺯﻴﻥ ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﻨﺎﺘﺠﺔ Saveﻭﺫﻟﻙ ﻜﻤﺎ ﻓﻲ ﺍﻟﺸﻜل 10-15ﺃﺩﻨﺎﻩ.
ﺸﻜل :10-15ﻨﺎﻓﺫﺓ ﺤﻔﻅ ﺍﻟﻨﺘﺎﺌﺞ Saveﻓﻲ ﻨﻤﺎﺫﺝ ARIMAﻟﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ.
( ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ15)
536
:ﻭﺘﻨﻔﻴﺫ ﻫﺫﺍ ﺍﻷﻤﺭ ﺴﻭﻑ ﻴﻌﻁﻲ ﺍﻟﻨﺘﺎﺌﺞ ﺍﻟﺘﺎﻟﻴﺔ MODEL:
MOD_2
Model Description: ﺠﻴﻨﻜﻨﺯ ﺍﻟﺫﻱ ﺘﻡ ﺘﻭﻓﻴﻘﻪ-ﻤﻌﻠﻭﻤﺎﺕ ﻋﺎﻤﺔ ﻋﻥ ﻨﻤﻭﺫﺝ ﺒﻭﻜﺱ Variable: PRICIND Regressors: NONE Non-seasonal differencing: 0 No seasonal component in model. Parameters: AR1 _______ <value originating from estimation > CONSTANT ________ <value originating from estimation > 95.00 percent confidence intervals will be generated. Split group number: 1 Series length: 48 No missing data. Melard's algorithm will be used for estimation.
: ﻤﻌﻠﻭﻤﺎﺕ ﻋﻥ ﺍﻟﻤﺭﺍﺤل ﺍﻟﺘﻲ ﻤﺭ ﺒﻬﺎ ﺍﻟﺒﺭﻨﺎﻤﺞ ﻟﻠﻭﺼﻭل ﺇﻟﻰ ﺍﻟﻨﺘﺎﺌﺞ ﺍﻟﻨﻬﺎﻴﺔ Termination criteria: Parameter epsilon: .001 Maximum Marquardt constant: 1.00E+09 SSQ Percentage: .001 Maximum number of iterations: 10 Initial values: AR1 .89332 CONSTANT 116.1333 Marquardt constant = .001 Adjusted sum of squares = 266.21919 Iteration 1
Iteration History: Adj. Sum of Squares 251.40466
Marquardt Constant .00100000
Conclusion of estimation phase. Estimation terminated at iteration number 2 because: Sum of squares decreased by less than .001 percent.
537
( ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ15)
FINAL PARAMETERS:
( )ﺍﻟﺠﺎﻨﺏ ﺍﻷﻜﺜﺭ ﺃﻫﻤﻴﺔ: ﺍﻟﻨﺘﺎﺌﺞ ﺍﻟﻨﻬﺎﺌﻴﺔ
Number of residuals Standard error Log likelihood AIC SBC
48 2.2739019
ﺴﻴﺘﻡ ﺘﻭﻀﻴﺢ ﻫﺫﺍ ﺍﻟﻤﻘﺩﺍﺭ ﻻﺤﻘ ﹰﺎ
-107.85164 219.70328 223.44568
Analysis of Variance: DF Adj. Sum of Squares Residuals 46 251.40351 Variables in the Model: B SEB AR1 .96440 .0355705 CONSTANT 115.89946 6.3193339
ﺘﺤﻠﻴل ﺍﻟﺘﺒﺎﻴﻥ ﻟﻠﻨﻤﻭﺫﺝ Residual Variance 5.1706297
AR(1) ﻤﻌﺎﻟﻡ ﻨﻤﻭﺫﺝ T-RATIO APPROX.PROB. 27.112420 .0000000 18.340455 .0000000
Covariance Matrix: AR1 AR1 .00126526 Correlation Matrix: AR1 AR1 1.0000000 Regressor Covariance Matrix: CONSTANT CONSTANT 39.933981 Regressor Correlation Matrix: CONSTANT CONSTANT 1.0000000
: ﺍﻟﻤﺘﻐﻴﺭﺍﺕ ﺍﻟﺠﺩﻴﺩﺓ ﺍﻟﺘﻲ ﺘﻡ ﺇﻀﺎﻓﺘﻬﺎ ﺇﻟﻰ ﻤﺤﺭﺭ ﺍﻟﺒﻴﺎﻨﺎﺕ The following new variables are being created: Name Label FIT_1 Fit for PRICIND from ARIMA, MOD_2 CON ERR_1 Error for PRICIND from ARIMA, MOD_2 CON LCL_1 95% LCL for PRICIND from ARIMA, MOD_2 CON UCL_1 95% UCL for PRICIND from ARIMA, MOD_2 CON SEP_1 SE of fit for PRICIND from ARIMA, MOD_2 CON 12 new cases have been added.
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
538
ﻭﺒﺎﻟﻨﻅﺭ ﺇﻟﻰ ﻤﻌﺎﻟﻡ ﺍﻟﻨﻤﻭﺫﺝ ) AR(1ﺍﻟﺫﻱ ﺘﻡ ﺘﻭﻓﻴﻘﻪ ﻟﻠﺒﻴﺎﻨﺎﺕ ﻓﻲ ﻗﺎﺌﻤﺔ ﺍﻟﻨﺘﺎﺌﺞ ﻴﺘﻀﺢ ﺃﻥ ﻫﺫﺍ ﺍﻟﻨﻤﻭﺫﺝ ﻴﺄﺨﺫ ﺍﻟﺼﻭﺭﺓ ﺍﻟﺘﺎﻟﻴﺔ: X t = 115.89946 + 0.96440 X t −1 + et
ﻭﻫﺫﺍ ﺍﻟﻨﻤﻭﺫﺝ ﻫﻭ ﺍﻟﺫﻱ ﻴﻤﻜﻥ ﺍﺴﺘﺨﺩﺍﻤﻪ ﻓﻲ ﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ ،ﻭﻴﺩل ﻋﻠﻰ ﺃﻨﻪ
ﺒﻤﻌﺭﻓﺔ ﻗﻴﻤﺔ ﻭﺤﻴﺩﺓ ﻓﻲ ﺍﻟﺴﻠﺴﻠﺔ )ﺒﺎﺴﺘﺨﺩﺍﻡ ﻤﺜل ﻫﺫﺍ ﺍﻟﻨﻤﻭﺫﺝ( ﻨﺴﺘﻁﻴﻊ ﺍﻟﺘﻨﺒﺅ ﺒﺎﻟﻘﻴﻤﺔ ﺍﻟﺘﻲ ﺘﻠﻴﻬﺎ ﻤﺒﺎﺸﺭﺓ ﻭﻤﻥ ﺜﻡ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﺍﻟﻤﺘﺘﺎﻟﻴﺔ ﺒﻤﻌﺩل ﺨﻁﺄ ﺒﺴﻴﻁ ،ﻭﻟﻜﻥ ﻗﺒل
ﺫﻟﻙ ﻴﻨﺒﻐﻲ ﺍﻟﺘﺤﻘﻕ ﻤﻥ ﻤﻼﺌﻤﺔ ﻫﺫﺍ ﺍﻟﻨﻤﻭﺫﺝ ﻟﺒﻴﺎﻨﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ،ﻭﺒﺎﻟﻨﻅﺭ ﺇﻟﻰ
ﺍﻟﻨﺘﺎﺌﺞ ﺍﻟﺴﺎﺒﻘﺔ ﻴﺘﻀﺢ ﺃﻨﻪ ﻻ ﻴﻭﺠﺩ ﻤﺸﻜﻠﺔ ﻭﺍﻀﺤﺔ ﺘﺩل ﻋﻠﻰ ﻋﺩﻡ ﻤﻼﺌﻤﺔ ﺍﻟﻨﻤﻭﺫﺝ
ﻟﻠﺒﻴﺎﻨﺎﺕ ،ﻭﻟﻜﻥ ﻟﻠﺘﺄﻜﺩ ﻤﻥ ﺫﻟﻙ ﻻ ﺒﺩ ﻤﻥ ﺍﻟﺘﺤﻘﻕ ﻤﻥ ﺍﻷﺨﻁﺎﺀ residualsﺒﺤﺜﹰﺎ ﻋﻥ
ﺃﻱ ﺍﻨﺘﻬﺎﻙ ﻟﻠﺸﺭﻭﻁ ﺍﻟﻤﻁﻠﻭﺒﺔ ﻟﺼﺤﺔ ﻫﺫﺍ ﺍﻟﻨﻤﻭﺫﺝ ،ﻭﺃﻫﻡ ﻫﺫﻩ ﺍﻟﺸﺭﻭﻁ ﻫﻭ ﺃﻥ ﻫﺫﻩ
ﺍﻷﺨﻁﺎﺀ ﻴﺠﺏ ﺃﻥ ﺘﻜﻭﻥ ﻏﻴﺭ ﻤﺘﺭﺍﺒﻁﺔ )ﺃﻱ ﻻ ﻴﻅﻬﺭ ﺒﻬﺎ ﺃﻱ ﻨﻤﻁ ﻤﻌﻴﻥ ﻴﺩل ﻋﻥ ﺃﻨﻬﺎ
ﻤﺘﺭﺍﺒﻁﺔ ﺫﺍﺘﻴﹰﺎ ،(autocorrelatedﻭﻓﻲ ﻤﺜل ﻫﺫﻩ ﺍﻟﺤﺎﻟﺔ ﻴﺤﺴﻥ ﺍﻟﺘﺤﻘﻕ ﻤﻥ ﻋﺩﻡ
ﻭﺠﻭﺩ ﻤﻌﺎﻤﻼﺕ ﻤﻌﻨﻭﻴﺔ ﻓﻲ ﺤﺩﻭﺩ ﺜﻘﺔ 95%ﻓﻲ ﻜل ﻤﻥ ﺩﺍﻟﺘﻲ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ACF
ﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ . PACF
ﻭﻴﻤﻜﻥ ﺭﺴﻡ ﺩﺍﻟﺘﻲ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ACFﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ PACF
ﻟﺴﻠﺴﻠﺔ ﺍﻷﺨﻁﺎﺀ ﺒﻨﻔﺱ ﻁﺭﻴﻘﺔ ﺭﺴﻤﻬﻤﺎ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻷﺼﻠﻴﺔ )ﻜﻤﺎ ﻓﻲ ﺍﻟﺸﻜل (6-15ﻋﻥ
ﻁﺭﻴﻕ ﺃﻤﺭ ﺭﺴﻡ ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ Time Series Plotﻤﻥ ﻗﺎﺌﻤﺔ ﺍﻟﺭﺴﻭﻤﺎﺕ ﺍﻟﺒﻴﺎﻨﻴﺔ Graphs
ﻓﻲ
ﺍﻟﻘﺎﺌﻤﺔ
ﺍﻟﺭﺌﻴﺴﻴﺔ
ﻟﻠﻨﻅﺎﻡ
ﻭﺍﺨﺘﻴﺎﺭ
ﺃﻤﺭ
ﺍﻟﺘﺭﺍﺒﻁ
ﺍﻟﺫﺍﺘﻲ
، Autocorrelationsﻭﻟﻜﻥ ﺍﻟﻤﺘﻐﻴﺭ ﺍﻟﺫﻱ ﻴﺠﺏ ﺃﻥ ﻴﺩﺨل ﻓﻲ ﻨﺎﻓﺫﺓ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ Autocorrelationsﻟﻠﺭﺴﻡ ﻓﻲ ﻫﺫﻩ ﺍﻟﺤﺎﻟﺔ ﻫﻭ ﻤﺘﻐﻴﺭ ﺍﻷﺨﻁﺎﺀ Residualsﺍﻟﺫﻱ ﺘﻡ ﺤﺴﺎﺒﻪ ﻓﻲ ﺍﻟﺨﻁﻭﺓ ﺍﻟﺴﺎﺒﻘﺔ ﻭﺘﻡ ﺘﺨﺯﻴﻨﻪ ﺒﺎﺴﻡ ، ERR_1ﻭﺒﺘﻁﺒﻴﻕ ﺫﻟﻙ ﻓﻲ ﻫﺫﺍ ﺍﻟﻤﺜﺎل
ﻓﺈﻨﻨﺎ ﺴﻨﺤﺼل ﻋﻠﻰ ﺭﺴﻡ ﻟﻨﻤﻁ ﻫﺎﺘﻴﻥ ﺍﻟﺩﺍﻟﺘﻴﻥ ﻜﻤﺎ ﻓﻲ ﺸﻜل 11-15ﺃﺩﻨﺎﻩ.
( ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ15)
539
. ﻟﻸﺨﻁﺎﺀPACF ﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲACF ﺩﺍﻟﺘﻲ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ: 11-15 ﺸﻜل ACF
Error for PRICIND from ARIMA, MOD_2 CON 1.0
.5
0.0
-.5
Confidence Limits
Coefficient
-1.0 1
3 2
5 4
7 6
9 8
11 10
13 12
15 14
16
Lag Number
Partial ACF
Error for PRICIND from ARIMA, MOD_2 CON 1.0
.5
0.0
-.5
Confidence Limits
Coefficient
-1.0 1
3 2
5 4
7 6
9 8
11 10
13 12
15 14
16
Lag Number
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
540
ﻴﺘﻀﺢ ﺒﺎﻟﻨﻅﺭ ﺇﻟﻰ ﺭﺴﻡ ﺩﺍﻟﺘﻲ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ACFﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ PACFﻟﺴﻠﺴﻠﺔ ﺍﻷﺨﻁﺎﺀ ﻓﻲ ﺸﻜل 11-15ﺍﻟﺴﺎﺒﻕ ﺃﻨﻪ ﻻ ﻴﻭﺠﺩ ﺃﻱ ﺍﺨﺘﺭﺍﻕ ﻟﻔﺭﻭﺽ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺫﻱ ﺘﻡ ﺘﻭﻓﻴﻘﻪ ﻟﻠﺒﻴﺎﻨﺎﺕ ،ﻭﺒﺫﻟﻙ ﻴﻤﻜﻨﻨﺎ ﺍﻟﺤﻜﻡ ﻋﻠﻰ ﺃﻥ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺫﻱ ﺘﻡ
ﺘﻭﻓﻴﻘﻪ ﻤﻼﺌﻡ ﻟﺒﻴﺎﻨﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﻭﺒﺎﻟﺘﺎﻟﻲ ﻓﺈﻨﻪ ﻴﻤﻜﻨﻨﺎ ﺃﻥ ﻨﻌﺘﻤﺩ ﻋﻠﻰ ﻫﺫﺍ ﺍﻟﻨﻤﻭﺫﺝ ﻓﻲ ﺍﻟﺘﻨﺒﺅ ﺒﻘﻴﻡ ﻤﺴﺘﻘﺒﻠﻴﺔ.
ﺸﻜل : 12-15ﺭﺴﻡ ﻟﺒﻴﺎﻨﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺍﻷﺼﻠﻴﺔ ﻭﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﻤﺘﻭﻗﻌﺔ ﺒﺎﺴﺘﺨﺩﺍﻡ ﺍﻟﻨﻤﻭﺫﺝ ) AR(1ﺍﻟﺫﻱ ﺘﻡ ﺘﻭﻓﻴﻘﻪ ﻭﺤﺩﻭﺩ ﺍﻟﺜﻘﺔ ﻟﻠﻘﻴﻡ ﺍﻟﻤﺘﻭﻗﻌﺔ. Observed monthly consumer price index, predicted values and 95% confidence interval of the predicted values.
130 Monthly Consumer Pri ce Index
120
Fit for PRICIND from ARIMA, MOD_2 CON
110
95% LCL for PRICIND from ARIMA, MOD_2 CO
100
95% UCL for PRICIND from ARIMA, MOD_2 CO 2001
Year
2000
1999
1998
90 1997
Consumer Price Index
140
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
541
ﻭﺃﺨﻴﺭﺍﹰ ،ﻓﻘﺩ ﺘﻡ ﺇﻋﻁﺎﺀ ﺍﻟﻨﻅﺎﻡ ﻓﻲ ﻫﺫﺍ ﺍﻟﻤﺜﺎل ﺃﻤﺭ ﺒﺎﻟﺘﻨﺒﺅ ﺒِـ 12ﻗﻴﻤﺔ ﻤﺴﺘﻘﺒﻠﻴﺔ ﻟﺴﻠﺴﻠﺔ ﺍﻷﺭﻗﺎﻡ ﺍﻟﻘﻴﺎﺴﻴﺔ ﺍﻟﺸﻬﺭﻴﺔ ﻟﻠﻤﺴﺘﻬﻠﻙ ﻓﻲ ﻓﻠﺴﻁﻴﻥ )ﺃﻱ ﻟﻠﻌﺎﻡ 2001
ﺒﻜﺎﻤﻠﻪ( ،ﻭﻜﻤﺎ ﻴﺘﻀﺢ ﻤﻥ ﻗﺎﺌﻤﺔ ﺍﻟﻨﺘﺎﺌﺞ ﺍﻟﺴﺎﺒﻘﺔ ﻓﺈﻥ ﻫﺫﻩ ﺍﻟﻘﻴﻡ ﺘﻅﻬﺭ ﻋﺎﺩﺓ ﻓﻲ ﺼﻔﺤﺔ
ﻤﺤﺭﺭ ﺍﻟﺒﻴﺎﻨﺎﺕ ﻓﻲ ﻋﻤﻭﺩ ﻤﺴﺘﻘل ﺒﺎﺴﻡ ،FIT_1ﻜﻤﺎ ﺃﻨﻪ ﻴﻤﻜﻥ ﺘﻭﻀﻴﺢ ﻫﺫﻩ ﺍﻟﻘﻴﻡ
ﺒﺎﻟﺭﺴﻡ ﺇﻀﺎﻓﺔ ﺇﻟﻰ ﺠﻤﻴﻊ ﺍﻟﻘﻴﻡ ﺍﻟﻤﻘﺩﺭﺓ ﺒﻭﺍﺴﻁﺔ ﺍﻟﻨﻤﻭﺫﺝ ﻓﻲ ﺸﻜل ﻭﺍﺤﺩ ﻤﻊ ﺍﻟﻘﻴﻡ
ﺍﻷﺼﻠﻴﺔ ﻟﻠﺴﻠﺴﻠﺔ ﻭﺤﺩﻱ ﺍﻟﺜﻘﺔ ﻟﻠﻘﻴﻡ ﺍﻟﻤﻘﺩﺭﺓ ﻭﺍﻟﻤﺘﻨﺒﺄ ﺒﻬﺎ ﻜﻤﺎ ﻓﻲ ﺸﻜل 12-15ﺃﻋﻼﻩ، ﻭﺭﻏﻡ ﺍﺨﺘﻔﺎﺀ ﺍﻷﻟﻭﺍﻥ ﻓﻲ ﺍﻟﻁﺒﺎﻋﺔ ﺇﻻ ﺃﻨﻪ ﻴﻤﻜﻨﻨﺎ ﺃﻥ ﻨﻤﻴﺯ ﺒﻴﻥ ﺍﻟﺴﻼﺴل ﺍﻷﺭﺒﻊ
ﺒﺴﻬﻭﻟﺔ ،ﺤﻴﺙ ﺘﻅﻬﺭ ﺤﺩﻭﺩ ﺍﻟﺜﻘﺔ ﺍﻷﺩﻨﻰ ﻭﺍﻷﻋﻠﻰ ﻟﻠﻘﻴﻡ ﺍﻟﻤﺘﻭﻗﻌﺔ ﻓﻲ ﺨﻁﻭﻁ ﻜل ﻤﻥ
ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺴﻔﻠﻴﺔ ﻭﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﻌﻠﻭﻴﺔ ،ﻴﺘﻭﺴﻁﻬﻤﺎ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺍﻷﺼﻠﻴﺔ )ﺍﻟﻘﺼﻴﺭﺓ(
ﺒﺎﻹﻀﺎﻓﺔ ﺇﻟﻰ ﺴﻠﺴﻠﺔ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺘﻭﻗﻌﺔ ﺍﻟﻤﻤﺘﺩﺓ ﺇﻟﻰ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺴﺘﻘﺒﻠﻴﺔ ﺍﻟﻤﺘﻨﺒﺄ ﺒﻬﺎ ،ﻻﺤﻅ ﺃﻥ ﻫﺎﺘﻴﻥ ﺍﻟﺴﻠﺴﻠﺘﻴﻥ ﻤﺘﻘﺎﺭﺒﺘﻴﻥ ﻓﻲ ﺍﻟﻘﻴﻡ ﻤﻤﺎ ﻴﺩل ﻋﻠﻰ ﺩﻗﺔ ﺘﻤﺜﻴل ﺍﻟﻨﻤﻭﺫﺝ ﻟﻠﺒﻴﺎﻨﺎﺕ. ﻭﻫﻨﺎﻙ ﺒﻌﺽ ﺍﻟﻤﻼﺤﻅﺎﺕ ﻋﻠﻰ ﻫﺫﺍ ﺍﻟﻤﺜﺎل ﻨﻭﺭﺩﻫﺎ ﻓﻴﻤﺎ ﻴﻠﻲ:
.1ﻟﻘﺩ ﺍﺨﺘﻴﺭﺕ ﻫﺫﻩ ﺍﻟﺴﻠﺴﻠﺔ ﻓﻲ ﻫﺫﺍ ﺍﻟﻤﺜﺎل ﺒﺎﻟﺘﺤﺩﻴﺩ ﻫﻨﺎ ﻷﻨﻬﺎ ﺴﻠﺴﻠﺔ ﺯﻤﻨﻴﺔ ﺤﻘﻴﻘﻴﺔ ﻭﻨﻅﺭﹸﺍ ﻹﻤﻜﺎﻨﻴﺔ ﺍﺴﺘﺨﺩﺍﻤﻬﺎ ﻓﻲ ﺘﻭﻀﻴﺢ ﺨﻁﻭﺍﺕ ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﺒﺴﻬﻭﻟﺔ
ﻭﻭﻀﻭﺡ ،ﻭﻗﺩ ﺘﺒﻴﻥ ﺃﻥ ﻤﻌﻅﻡ ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﺍﻟﻤﺘﻌﻠﻘﺔ ﺒﺎﻷﺭﻗﺎﻡ ﺍﻟﻘﻴﺎﺴﻴﺔ
ﺍﻟﺸﻬﺭﻴﺔ ﻟﻠﻤﺴﺘﻬﻠﻙ ﻓﻲ ﺍﻟﻜﺜﻴﺭ ﻤﻥ ﺍﻟﺒﻠﺩﺍﻥ ﺘﺄﺨﺫ ﻨﻔﺱ ﻨﻤﻁ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺴﺎﺒﻕ ﻭﻴﻤﻜﻥ
ﺃﻥ ﻴﺴﺘﺨﺩﻡ ﻫﺫﺍ ﺍﻟﻨﻤﻭﺫﺝ )ﺍﻨﻅﺭ ﻋﻠﻰ ﺴﺒﻴل ﺍﻟﻤﺜﺎل Bowerman & O'Connell,
، (1993ﻭﻴﻤﻜﻥ ﺍﻟﺘﺩﺭﺏ ﻋﻠﻰ ﺒﻴﺎﻨﺎﺕ ﺍﻟﻌﺩﻴﺩ ﻤﻥ ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﺍﻟﻤﺘﻭﻓﺭﺓ ﻤﻊ ﻨﻅﺎﻡ SPSSﺫﺍﺘﻪ ﻓﻲ ﻤﻠﻑ ﺍﻟﺒﻴﺎﻨﺎﺕ .Data
.2ﻴﺘﻀﺢ ﻤﻥ ﻫﺫﺍ ﺍﻟﻤﺜﺎل ﺃﻥ ﺍﻟﺨﻁﻭﺓ ﺍﻷﻭﻟﻰ ﺍﻟﻬﺎﻤﺔ ﻓﻲ ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﺒﻌﺩ
ﺭﺴﻡ ﺍﻟﺴﻠﺴﻠﺔ ﻫﻲ ﺘﺤﺩﻴﺩ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﻤﺒﺩﺌﻲ ﻤﻥ ﺨﻼل ﻓﺤﺹ ﺩﺍﻟﺘﻲ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ
ACFﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ ، PACFﻭﻤﻥ ﺨﻼﻟﻬﻤﺎ ﻴﻤﻜﻨﻨﺎ ﺒﺴﻬﻭﻟﺔ ﺍﻗﺘﺭﺍﺡ ﻨﻤﻭﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ARIMAﻟﻨﺒﺩﺃ ﺒﻪ.
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
542
.3ﻓﻲ ﻗﺎﺌﻤﺔ ﺍﻟﻨﺘﺎﺌﺞ ﻓﻲ ﻨﻅﺎﻡ SPSSﺘﻅﻬﺭ ﻨﺘﺎﺌﺞ ﻜﺜﻴﺭﺓ ﻤﻥ ﺒﻴﻨﻬﺎ ﺍﻟﺘﻘﺩﻴﺭﺍﺕ ﺍﻟﻤﺒﺩﺌﻴﺔ ﻟﻠﻨﻤﻭﺫﺝ ﺍﻟﻤﻁﻠﻭﺏ ﻭﻤﺭﺍﺤل ﺘﻁﻭﺭﻩ ،ﻭﻴﻤﻜﻥ ﺘﺠﺎﻫل ﻜل ﻫﺫﻩ ﺍﻟﻤﻌﻠﻭﻤﺎﺕ ﻭﺍﻻﻫﺘﻤﺎﻡ ﻓﻘﻁ ﺒﺎﻟﻨﺘﺎﺌﺞ ﺍﻟﻨﻬﺎﺌﻴﺔ . FINAL ESTIMATES
.4ﻫﻨﺎﻙ ﺒﻌﺽ ﺍﻟﻤﻘﺎﻴﻴﺱ ﺍﻟﻬﺎﻤﺔ ﻭﺍﻟﺘﻲ ﺘﻅﻬﺭ ﻓﻲ ﻗﺎﺌﻤﺔ ﺍﻟﻨﺘﺎﺌﺞ ﻭﻴﻤﻜﻥ ﺍﺴﺘﺨﺩﺍﻤﻬﺎ ﻓﻲ ﺍﻟﺘﺤﻘﻕ ﻤﻥ ﻤﺩﻯ ﻤﻼﺌﻤﺔ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺫﻱ ﺘﻡ ﺘﻭﻓﻴﻘﻪ ﻟﻠﺴﻠﺴﻠﺔ ،ﻫﺫﻩ ﺍﻟﻤﻘﺎﻴﻴﺱ ﻫﻲ
ﻤﺠﻤﻭﻉ ﻤﺭﺒﻌﺎﺕ ﺍﻷﺨﻁﺎﺀ residual sum of squaresﻭﻟﻭﻏﺎﺭﻴﺘﻡ ﺩﺍﻟﺔ
ﺍﻷﺭﺠﺤﻴﺔ log likelihood functionﻓﻜﻠﻤﺎ ﻜﺎﻨﺕ ﻗﻴﻤﺘﻲ ﻫﺫﻴﻥ ﺍﻟﻤﻘﻴﺎﺴﻴﻥ ﺼﻐﻴﺭﺘﻴﻥ ﻜﻠﻤﺎ ﻜﺎﻥ ﺍﻟﻨﻤﻭﺫﺝ ﺃﺠﻭﺩ ﻟﻠﺒﻴﺎﻨﺎﺕ ،ﻭﺒﻌﺩ ﺫﻟﻙ ﻴﻤﻜﻥ ﺍﻟﻨﻅﺭ ﺇﻟﻰ ﻤﻌﻴﺎﺭ
ﻤﻌﻠﻭﻤﺎﺕ ﺃﻜﺎﻜﻲ ) ، Akaike’s Information Criterion (AICﻭﻤﻌﻴﺎﺭ
ﻤﻌﻠﻭﻤﺎﺕ ﺸﻴﻭﺍﺭﺯ ﺍﻟﻤﻌﺘﻤﺩ ﻋﻠﻰ ﺃﺴﻠﻭﺏ ﺒﻴﺯ Schwarz Bayesian Information
) ، Criterion (SBCﻭﻴﻌﺘﻤﺩ ﺤﺴﺎﺏ ﻫﺫﻴﻥ ﺍﻟﻤﻌﻴﺎﺭﻴﻥ ﻋﻠﻰ ﻗﻴﻤﺔ ﻤﺭﺒﻊ ﻤﻌﺎﻤل ﺍﻻﺭﺘﺒﺎﻁ ، R2ﺇﻻ ﺃﻨﻬﻤﺎ ﻴﺄﺨﺫﺍﻥ ﻓﻲ ﺍﻻﻋﺘﺒﺎﺭ ﻋﻭﺍﻤل ﺃﺨﺭﻯ ﺇﻀﺎﻓﻴﺔ ،ﻭﻟﻜﻥ ﻴﺠﺏ
ﺍﻷﺨﺫ ﻓﻲ ﺍﻻﻋﺘﺒﺎﺭ ﺃﻥ ﻫﺫﻴﻥ ﺍﻟﻤﻌﻴﺎﺭﻴﻥ ﻨﺴﺒﻴﻴﻥ ﻭﻴﻌﺘﻤﺩﺍﻥ ﻋﻠﻰ ﻗﻴﻡ ﺍﻟﺒﻴﺎﻨﺎﺕ ﻭﻟﺫﺍ ﻴﻤﻜﻥ ﺍﺴﺘﺨﺩﺍﻤﻬﻤﺎ ﻓﻘﻁ ﻓﻲ ﺍﻟﻤﻘﺎﺭﻨﺔ ﺒﻴﻥ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﻤﺨﺘﻠﻔﺔ ﻟﻨﻔﺱ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ، ﻭﺍﻟﻨﻤﻭﺫﺝ ﺍﻷﻓﻀل ﻫﻭ ﺍﻟﺫﻱ ﻴﺼﺎﺤﺒﻪ ﻗﻴﻤﺘﻴﻥ ﺃﺼﻐﺭ ﻟﻬﺫﻴﻥ ﺍﻟﻤﻌﻴﺎﺭﻴﻥ .
.5ﻻﺒﺩ ﺃﻴﻀﹰﺎ ﻤﻥ ﻓﺤﺹ ﻤﻌﻨﻭﻴﺔ ﻤﻌﺎﻤﻼﺕ ﺍﻟﻨﻤﻭﺫﺝ ﻗﺒل ﺇﺼﺩﺍﺭ ﺍﻟﺤﻜﻡ ﺍﻟﻨﻬﺎﺌﻲ ﻋﻠﻴﻪ،
ﻓﻴﺠﺏ ﺃﻻ ﻴﻜﻭﻥ ﺃﻱ ﻤﻥ ﻫﺫﻩ ﺍﻟﻤﻌﺎﻤﻼﺕ ﻤﺴﺎﻭﻴﹰﺎ ﻤﻌﻨﻭﻴﹰﺎ ﻟﻠﺼﻔﺭ ﻭﺇﻻ ﻴﺠﺏ ﺤﺫﻓﻪ ﻤﻥ ﺍﻟﻨﻤﻭﺫﺝ ،ﻭﻴﻤﻜﻥ ﺍﺨﺘﺒﺎﺭ ﻤﻌﻨﻭﻴﺔ ﺍﻟﻤﻌﺎﻤﻼﺕ ﺒﺎﺴﺘﺨﺩﺍﻡ ﻗﻴﻤﺔ p-value
ﺍﻟﻤﺭﺍﻓﻘﺔ ﻟﻜل ﻤﻌﺎﻤل ﻓﻲ ﻗﺎﺌﻤﺔ ﺍﻟﻨﺘﺎﺌﺞ.
.6ﻴﺠﺏ ﻓﺤﺹ ﺍﻷﺨﻁﺎﺀ residualsﺃﻴﻀﹰﺎ ﻋﻥ ﻁﺭﻴﻕ ﺭﺴﻤﻬﺎ ﻓﻲ ﺸﻜل ﺍﻨﺘﺸﺎﺭ ﻤﻘﺎﺒل ﺍﻟﻘﻴﻡ ﺍﻟﻤﺘﻭﻗﻌﺔ fitted valueﺒﺎﻹﻀﺎﻓﺔ ﺇﻟﻰ ﺭﺴﻡ ﺩﺍﻟﺘﻲ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ
ACFﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ PACFﻟﻸﺨﻁﺎﺀ ﻭﺍﻟﻠﺘﻴﻥ ﻴﺠﺏ ﺃﻥ ﺘﻅﻬﺭﺍ ﻋﺩﻡ ﻭﺠﻭﺩ ﻤﻌﺎﻤﻼﺕ ﻤﻌﻨﻭﻴﺔ ﻓﻲ ﻫﺎﺘﻴﻥ ﺍﻟﺩﺍﻟﺘﻴﻥ.
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
543
.7ﻴﻔﻀل ﺩﺍﺌﻤﹰﺎ ﺘﺠﺭﺒﺔ ﺃﻜﺜﺭ ﻤﻥ ﻨﻤﻭﺫﺝ ﻭﺍﺤﺩ ﻤﻥ ﻨﻤﺎﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ARIMA ﻭﺫﻟﻙ ﺒﺄﺨﺫ ﻨﻤﺎﺫﺝ ﺒﻤﻌﺎﻤﻼﺕ ﺇﻀﺎﻓﻴﺔ ﺠﺩﻴﺩﺓ ﻭ ﻨﻤﺎﺫﺝ ﻨﺘﺠﺕ ﻤﻥ ﺤﺫﻑ ﻤﻌﺎﻤﻼﺕ
ﻤﻭﺠﻭﺩﺓ ﻓﻲ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻷﺼﻠﻲ ﻭﺍﻟﻤﻘﺎﺭﻨﺔ ﺒﻴﻥ ﺠﻤﻴﻊ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﻨﺎﺘﺠﺔ ،ﻭﺒﺎﻟﺘﺎﻟﻲ
ﻻﺒﺩ ﻤﻥ ﺍﺨﺘﻴﺎﺭ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻷﻜﺜﺭ ﺠﻭﺩﺓ ﺒﻨﺎﺀ ﻋﻠﻰ ﺍﻟﻤﻌﺎﻴﻴﺭ ﺍﻟﺴﺎﺒﻘﺔ.
ﻭﺭﻏﻡ ﺃﻥ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺴﺎﺒﻕ ﻴﺒﺩﻭ ﻤﻼﺌﻡ ﻟﺒﻴﺎﻨﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺇﻻ ﺃﻨﻪ ﺒﺴﺒﺏ
ﻭﺠﻭﺩ ﺍﻟﺘﻐﻴﺭ ﺍﻟﻤﻔﺎﺠﺊ ﻓﻲ ﺍﻟﺴﻠﺴﻠﺔ Step Changeﻴﻤﻜﻥ ﺘﺤﺴﻴﻥ ﺍﻟﻨﻤﻭﺫﺝ ﻭﺍﻟﺤﺼﻭل
ﻋﻠﻰ ﻨﻤﻭﺫﺝ ﺃﻓﻀل ﻤﻥ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺴﺎﺒﻕ ﻋﻥ ﻁﺭﻴﻕ ﺇﺩﺨﺎل ﻤﺎ ﻴﻁﻠﻕ ﻋﻠﻴﻪ ﺒﺎﻟﻤﺘﻐﻴﺭ ﺍﻻﺼﻁﻨﺎﻋﻲ Dummy Variableﻋﻨﺩ ﺍﻟﻘﻴﻤﺔ ﺭﻗﻡ 26ﻓﻲ ﺍﻟﺴﻠﺴﻠﺔ ،ﻭﻴﻤﻜﻥ ﺇﻀﺎﻓﺔ
ﻫﺫﺍ ﺍﻟﻤﺘﻐﻴﺭ ﺍﻻﺼﻁﻨﺎﻋﻲ ﺇﻟﻰ ﻤﺤﺭﺭ ﺍﻟﺒﻴﺎﻨﺎﺕ ﻋﻥ ﻁﺭﻴﻕ ﺃﻤﺭ ﺍﻟﺘﺼﻨﻴﻑ Recodeﻓﻲ
ﻗﺎﺌﻤﺔ ﺘﺤﻭﻴل ﺍﻟﻤﺘﻐﻴﺭﺍﺕ Transformﻓﻲ ﺍﻟﻘﺎﺌﻤﺔ ﺍﻟﺭﺌﻴﺴﻴﺔ ﻟﻤﺤﺭﺭ ﺍﻟﺒﻴﺎﻨﺎﺕ ،ﺜﻡ ﻴﺘﻡ
ﺇﺩﺨﺎل ﻫﺫﺍ ﺍﻟﻤﺘﻐﻴﺭ ﻓﻲ ﺘﺤﻠﻴل ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﻜﻤﺘﻐﻴﺭ ﺘﻌﺭﻴﻑ Indicator Variable
ﻋﻨﺩ ﺘﻭﻓﻴﻕ ﻨﻤﻭﺫﺝ ). AR(1
.6 .15ﺗﺤﻠﻴﻞ ﻧﻤﺎذج اﻟﺴﻼﺳﻞ اﻟﺰﻣﻨﻴﺔ ﻏﻴﺮ اﻟﺴﺎآﻨﺔ : Analysis of Non Stationary Series : ﺇﺫﺍ ﺘﺒﻴﻥ ﺃﻥ ﻫﻨﺎﻙ ﺃﺜﺭ ﻭﺍﻀﺢ ﻟﻼﺘﺠﺎﻩ ﺍﻟﻌﺎﻡ Trendﻓﻲ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ،
ﺍﻷﻤﺭ ﺍﻟﺫﻱ ﻴﻌﻠﻥ ﻋﻨﻪ ﺍﻟﺭﺴﻡ ﻓﻲ ﺸﻜل ﺍﻟﺴﻠﺴﻠﺔ Sequence Plotﺃﻭ ﻴﻭﻀﺤﻪ ﺸﻜل
ﺍﻟﺼﻨﺎﺩﻴﻕ Box Plotﺃﻭ ﻗﻴﻡ ﻤﻌﺎﻤﻼﺕ ﺩﺍﻟﺔ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ PACF
ﺍﻟﻤﺘﻌﺎﻅﻤﺔ ﻓﺈﻨﻪ ﻻﺒﺩ ﻤﻥ ﺤﺴﺎﺏ ﺴﻠﺴﻠﺔ ﺍﻟﻔﺭﻭﻕ ﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺍﻷﺼﻠﻴﺔ ،ﻭﺍﻟﺘﻲ ﻴﻤﻜﻥ ﺍﻟﺘﻌﺒﻴﺭ ﻋﻨﻬﺎ ﺒﺎﻟﻨﻤﻭﺫﺝ ، Zt = Xt - Xt-1ﻭﻴﻤﻜﻥ ﺍﻟﺘﻌﺒﻴﺭ ﻋﻥ ﻨﻤﻭﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ
ﻓﻲ ﻫﺫﺍ ﺍﻟﺤﺎﻟﺔ ﺒﺎﻟﻨﻤﻭﺫﺝ ) ، ARIMA(p,I,qﺤﻴﺙ Iﻫﻲ ﻋﺩﺩ ﻤﺭﺍﺕ ﺃﺨﺫ ﺍﻟﻔﺭﻕ ﻭﻫﻲ ﺍﺨﺘﺼﺎﺭﹰﺍ ﻟﻜﻠﻤﺔ ﺍﻟﻤﺘﻜﺎﻤل Integratedﻟﻨﻤﻭﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ .
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
544
ﻭﻟﺘﻭﻀﻴﺢ ﻜﻴﻔﻴﺔ ﺍﻟﺘﻌﺎﻤل ﻓﻲ ﻨﻅﺎﻡ SPSSﻤﻊ ﻨﻤﺎﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﺍﻟﻤﺘﻜﺎﻤﻠﺔ ARIMAﻭﺍﻟﺘﻲ ﺘﺤﺘﻭﻱ ﻋﻠﻰ ﺃﺜﺭ ﻟﻼﺘﺠﺎﻩ ﺍﻟﻌﺎﻡ ﺴﻨﺘﻨﺎﻭل ﺍﻟﻤﺜﺎل ﺍﻟﺘﺎﻟﻲ ﻭﺍﻟﺫﻱ ﻴﺘﻀﻤﻥ ﺒﻴﺎﻨﺎﺕ ﺤﻘﻴﻘﻴﺔ ﺃﻴﻀﺎﹰ ،ﻭﻫﺫﻩ ﺍﻟﺴﻠﺴﻠﺔ ﺘﻡ ﺘﺤﻠﻴﻠﻬﺎ ﻭﻨﺸﺭ ﻨﺘﺎﺌﺞ ﺍﻟﺘﺤﻠﻴل ﻓﻲ ﺍﻟﻌﺩﻴﺩ ﻤﻥ
ﺍﻟﻜﺘﺏ ﻭﺍﻷﺒﺤﺎﺙ ﺍﻟﻌﻠﻤﻴﺔ ﻭﺘﺘﻀﻤﻥ ﺃﻋﺩﺍﺩ ﺤﻘﻭل ﺍﻟﺒﺘﺭﻭل ﺍﻟﻨﺸﻁﺔ ﻓﻲ ﺍﻟﻭﻻﻴﺎﺕ ﺍﻟﻤﺘﺤﺩﺓ ﺍﻷﻤﺭﻴﻜﻴﺔ ﻓﻲ 50ﺸﻬﺭﹰﺍ ﻤﺘﺘﺎﻟﻴﺎﹰ ،ﻭﻟﻥ ﻨﻌﺭﺽ ﻫﻨﺎ ﺴﻠﺴﻠﺔ ﻟﻠﺒﻴﺎﻨﺎﺕ ﺒﻬﺩﻑ ﻋﺩﻡ ﺍﻟﺘﺭﻜﻴﺯ
ﻋﻠﻰ ﺍﻟﻨﻭﺍﺤﻲ ﺍﻹﺠﺭﺍﺌﻴﺔ ﻓﻲ ﻨﻅﺎﻡ SPSSﻓﻨﺄﻤل ﺃﻥ ﻴﻜﻭﻥ ﻫﺫﺍ ﻭﺍﻀﺤﹰﺎ ﻓﻲ ﻫﺫﻩ
ﺍﻟﻤﺭﺤﻠﺔ ،ﺒل ﻨﺭﻜﺯ ﺍﻻﻫﺘﻤﺎﻡ ﺍﻵﻥ ﻋﻠﻰ ﻁﺭﻴﻘﺔ ﺍﺨﺘﻴﺎﺭ ﻨﻤﻭﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ﺍﻟﻤﺘﻜﺎﻤل ARIMAﺍﻋﺘﻤﺎﺩﹰﺍ ﻋﻠﻰ ﺍﻟﺭﺴﻭﻤﺎﺕ ﺍﻟﺒﻴﺎﻨﻴﺔ ﺍﻟﺘﻲ ﻴﻤﻜﻨﻨﺎ ﺍﻟﻨﻅﺎﻡ ﻤﻥ ﺇﺠﺭﺍﺀﻫﺎ
ﺒﺴﻬﻭﻟﺔ ،ﻭﺍﻟﺘﺤﻠﻴل ﺍﻟﻤﺒﺩﺌﻲ ﻟﻬﺫﻩ ﺍﻟﺴﻠﺴﻠﺔ ﻴﻌﻁﻲ ﺍﻷﺸﻜﺎل ﺍﻟﺘﺎﻟﻴﺔ ﻟﻜل ﻤﻥ ﻨﻤﻁ ﺍﻟﺘﻐﻴﺭ
ﻓﻲ ﺍﻟﺴﻠﺴﻠﺔ )ﺸﻜل (13-15ﻭﺸﻜل ﺍﻟﺼﻨﺎﺩﻴﻕ ) Box plotﺸﻜل (14-15ﻜﻤﺎ ﻴﻠﻲ.
ﺸﻜل : 13-15ﺴﻠﺴﻠﺔ ﺯﻤﻨﻴﺔ ﺘﻤﺜل ﻋﺩﺩ ﺁﺒﺎﺭ ﺍﻟﺒﺘﺭﻭل ﺍﻟﻨﺸﻁﺔ Active Oil Wellsﻓﻲ 50ﺸﻬﺭﹰﺍ ﻓﻲ ﺍﻟﻭﻻﻴﺎﺕ ﺍﻟﻤﺘﺤﺩﺓ ﺍﻷﻤﺭﻴﻜﻴﺔ. 700000
600000
500000
400000
10 13 16 19 22 25 28 31 34 37 40 43 46 49
7
4
1
Sequence number
WELLS
300000
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
545
ﺸﻜل : 14-15ﺃﺸﻜﺎل ﺍﻟﺼﻨﺎﺩﻴﻕ Box Plotsﻟﺴﻠﺴﻠﺔ ﻋﺩﺩ ﺁﺒﺎﺭ ﺍﻟﺒﺘﺭﻭل ﺍﻟﻨﺸﻁﺔ Active Oil Wellsﻓﻲ 50ﺸﻬﺭﹰﺍ ﻓﻲ ﺍﻟﻭﻻﻴﺎﺕ ﺍﻟﻤﺘﺤﺩﺓ ﺍﻷﻤﺭﻴﻜﻴﺔ. 700000
50
600000
500000
400000
14
12
12
12
4.00
3.00
2.00
1.00
WELLS
300000 =N
VAR00002
ﻭﻴﺘﻀﺢ ﻤﻥ ﺸﻜل ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺍﻟﺴﺎﺒﻕ )ﺸﻜل (13-15ﻭﺠﻭﺩ ﺃﺜﺭ ﻟﻼﺘﺠﺎﻩ
ﺍﻟﻌﺎﻡ Trendﻭﺍﻟﺫﻱ ﻴﺅﻜﺩﻩ ﻨﻤﻁ ﺃﺸﻜﺎل ﺍﻟﺼﻨﺎﺩﻴﻕ )ﺸﻜل (14-15ﺍﻷﻤﺭ ﺍﻟﺫﻱ ﻴﺅﺩﻱ
ﺇﻟﻰ ﺍﻻﺴﺘﻨﺘﺎﺝ ﺒﻀﺭﻭﺭﺓ ﺤﺴﺎﺏ ﺴﻠﺴﻠﺔ ﺍﻟﻔﺭﻭﻕ Differencingﻤﻥ ﺍﻟﺴﻠﺴﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺍﻷﺼﻠﻴﺔ ،ﻭﻓﻲ ﻫﺫﻩ ﺍﻟﺤﺎﻟﺔ ﻨﺒﺩﺃ ﺒﺘﺠﺭﺒﺔ ﺍﻟﻔﺭﻭﻕ ﻋﻨﺩ ﻓﺘﺭﺓ lagﻭﺍﺤﺩﺓ ،ﻭﻟﻠﺘﺄﻜﺩ ﻤﻥ ﺃﻥ
ﻫﺫﻩ ﺍﻟﺘﺤﻭﻴﻠﺔ ﻗﺩ ﺃﻋﻁﺕ ﺍﻟﻨﺘﺎﺌﺞ ﺍﻟﻤﺭﺠﻭﺓ ﺒﺈﺯﺍﻟﺔ ﺍﻻﺘﺠﺎﻩ ﺍﻟﻌﺎﻡ Trendﻤﻥ ﺍﻟﺴﻠﺴﻠﺔ
ﺍﻟﺯﻤﻨﻴﺔ ﻴﺤﺴﻥ ﺃﻥ ﻨﺴﺘﻜﺸﻑ ﻨﻤﻁ ﻜل ﻤﻥ ﺩﺍﻟﺔ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ACFﻭﺩﺍﻟﺔ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ ) PACFﻭﺍﻟﺘﻴﻥ ﺘﻡ ﺭﺴﻤﻬﻤﺎ ﻓﻲ ﺸﻜل .(15-15
ﻭﻜﻤﺎ ﻴﺘﻀﺢ ﻤﻥ ﻨﻤﻁ ﺩﺍﻟﺔ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ACFﻭﺩﺍﻟﺔ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ
) PACFﺸﻜل (15-15ﻓﺈﻥ ﺍﻟﻌﻤﻠﻴﺔ ﺍﻟﺴﺎﺒﻘﺔ ﻟﻡ ﺘﺘﻤﻜﻥ ﻤﻥ ﺇﺯﺍﻟﺔ ﺃﺜﺭ ﺍﻻﺘﺠﺎﻩ ﺍﻟﻌﺎﻡ
ﻜﻠﻪ ﻤﻥ ﺍﻟﺴﻠﺴﻠﺔ ،ﻭﻫﺫﺍ ﻴﻘﺘﺭﺡ ﺘﻜﺭﺍﺭ ﺤﺴﺎﺏ ﺴﻠﺴﻠﺔ ﺍﻟﻔﺭﻭﻕ Differencingﻭﺭﺴﻡ
ﺩﺍﻟﺔ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ACFﻭﺩﺍﻟﺔ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ PACFﻟﻠﺴﻠﺴﻠﺔ ﺍﻟﺠﺩﻴﺩﺓ )ﻭﻓﻲ ﻫﺫﺍ ﺍﻟﻤﺜﺎل ﺸﻜل (16-15ﻟﺤﻴﻥ ﺍﻟﺤﺼﻭل ﻋﻠﻰ ﺴﻠﺴﻠﺔ ﺨﺎﻟﻴﺔ ﻤﻥ ﺍﻻﺘﺠﺎﻩ ﺍﻟﻌﺎﻡ.
( ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ15)
546
PACF ﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲACF ﻨﻤﻁ ﺩﺍﻟﺘﻲ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ: 15 -15 ﺸﻜل Active Oil Wells ﻟﺴﻠﺴﻠﺔ ﺍﻟﻔﺭﻭﻕ ﻤﺭﺓ ﻭﺍﺤﺩﺓ ﻤﻥ ﺴﻠﺴﻠﺔ ﻋﺩﺩ ﺁﺒﺎﺭ ﺍﻟﺒﺘﺭﻭل ﺍﻟﻨﺸﻁﺔ .ﻓﻲ ﺍﻟﻭﻻﻴﺎﺕ ﺍﻟﻤﺘﺤﺩﺓ ﺍﻷﻤﺭﻴﻜﻴﺔ
WELLS 1.0
.5
0.0
ACF
-.5
Confidence Limits
Coefficient
-1.0 1
3 2
5 4
7 6
9 8
11 10
13 12
15 14
16
Lag Number Transforms: difference (1)
WELLS 1.0
.5
Partial ACF
0.0
-.5
Confidence Limits
-1.0
Coefficient 1
3 2
5 4
7 6
Lag Number Transforms: difference (1)
9 8
11 10
13 12
15 14
16
( ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ15)
547
PACF ﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲACF ﻨﻤﻁ ﺩﺍﻟﺘﻲ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ: 16 -15 ﺸﻜل Active Oil ﻟﺴﻠﺴﻠﺔ ﺍﻟﻔﺭﻭﻕ ﻤﺭﺘﻴﻥ ﻤﺘﺘﺎﻟﻴﺘﻴﻥ ﻤﻥ ﺴﻠﺴﻠﺔ ﻋﺩﺩ ﺁﺒﺎﺭ ﺍﻟﺒﺘﺭﻭل ﺍﻟﻨﺸﻁﺔ . ﻓﻲ ﺍﻟﻭﻻﻴﺎﺕ ﺍﻟﻤﺘﺤﺩﺓ ﺍﻷﻤﺭﻴﻜﻴﺔWells
WELLS 1.0
.5
0.0
ACF
-.5
Confidence Limits
-1.0
Coefficient 1
3 2
5 4
7 6
9 8
11 10
13 12
15 14
16
Lag Number Transforms: difference (2)
WELLS 1.0
.5
Partial ACF
0.0
-.5
Confidence Limits
-1.0
Coefficient 1
3 2
5 4
7 6
Lag Number Transforms: difference (2)
9 8
11 10
13 12
15 14
16
548
( ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ15)
ﻓﻲ ﺍﻟﺸﻜل ﺍﻟﺴﺎﺒﻕ ﺘﺘﻭﻗﻑ ﻋﻨﺩ ﺍﻟﻔﺘﺭﺓACF ﻭﺤﻴﺙ ﺃﻥ ﺩﺍﻟﺔ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺘﺩﺭﻴﺠﻴﹰﺎ ﺘﻘﺭﻴﺒﹰﺎ ﻓﺈﻥ ﻫﺫﺍPACF ﺍﻟﺜﺎﻨﻴﺔ ﺒﻴﻨﻤﺎ ﺘﺘﻼﺸﻰ ﺩﺍﻟﺔ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ ﻜﻨﻤﻭﺫﺝ ﻤﺒﺩﺌﻲ ﻟﻠﺴﻠﺴﻠﺔARIMA (0,2,2) ﺠﻴﻨﻜﻨﺯ-ﻴﻘﺘﺭﺡ ﺃﺨﺫ ﻨﻤﻭﺫﺝ ﺒﻭﻜﺱ
. ﺍﻟﻨﺎﺘﺠﺔResiduals ﻟﺫﺍ ﺴﻭﻑ ﻨﻘﻭﻡ ﺒﺘﻭﻓﻴﻕ ﻫﺫﺍ ﺍﻟﻨﻤﻭﺫﺝ ﻭﻓﺤﺹ ﺍﻷﺨﻁﺎﺀ،ﺍﻟﺯﻤﻨﻴﺔ
ﻟﺒﻴﺎﻨﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺒﺎﺴﺘﺨﺩﺍﻡARIMA(0,2,2) ﻭﺒﺘﻭﻓﻴﻕ ﻫﺫﺍ ﺍﻟﻨﻤﻭﺫﺝ
: ﺴﻭﻑ ﻨﺤﺼل ﻋﻠﻰ ﺍﻟﻨﺘﺎﺌﺞ ﺍﻟﺘﺎﻟﻴﺔSPSS ﻨﻅﺎﻡ
Model Description: Variable: WELLS Regressors: NONE Non-seasonal differencing: 2 No seasonal component in model. 95.00 percent confidence intervals will be generated. Split group number: 1 Series length: 50 Number of cases skipped at end because of missing values: 10 Melard's algorithm will be used for estimation. Termination criteria: Parameter epsilon: .001 Maximum Marquardt constant: 1.00E+09 SSQ Percentage: .001 Maximum number of iterations: 10
Initial values: MA1 .63781 MA2 -.37588 CONSTANT 295.0479
Marquardt constant = .001 Adjusted sum of squares = 4661182951.2
( ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ15)
549
Iteration History: Iteration
Adj. Sum of Squares
Marquardt Constant
1 2 3 4 5 6
4441938081.4 4340471313.1 4323324857.5 4320922537.6 4320444357.3 4320360325.9
.00100000 .00010000 .00001000 .00000100 .00000010 .0000000
Conclusion of estimation phase. Estimation terminated at iteration number 7 because: Sum of squares decreased by less than .001 percent.
FINAL PARAMETERS: Number of residuals Standard error Log likelihood AIC SBC
48 9717.243 -507.70576 1021.4115 1027.0251
Analysis of Variance: DF Adj. Sum of Squares Residuals 45 4320343657.9
Variables in the Model: B SEB MA1 MA2 CONSTANT
.86031 .13937 -.37266 .14103 317.04973 721.98652
Residual Variance 94424811.9
T-RATIO
APPROX. PROB.
6.1728357 -2.6423294 .4391352
.00000016 .01128646 .66266502
ﻭﺩﺍﻟﺔACF ﺍﻟﺘﺎﻟﻲ ﻨﻤﻁ ﻜل ﻤﻥ ﺩﺍﻟﺔ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ17-15 ﻭﻴﻭﻀﺢ ﺸﻜل
ﻭﻫﺫﺍ ﺍﻟﺸﻜل ﻴﺒﻴﻥ ﺃﻥ،residuals ﻟﺴﻠﺴﻠﺔ ﺍﻷﺨﻁﺎﺀPACF ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲ
ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﻤﻘﺘﺭﺡ ﻤﻨﺎﺴﺏ ﻟﺒﻴﺎﻨﺎﺕ ﺘﻠﻙ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﻭﺫﻟﻙ ﻟﻌﺩﻡ ﻭﺠﻭﺩ ﻤﻌﺎﻤﻼﺕ . 95% ﻤﻌﻨﻭﻴﺔ ﻓﻲ ﻜل ﻤﻥ ﺍﻟﺩﺍﻟﺘﻴﻥ ﺒﺩﺭﺠﺔ ﺜﻘﺔ
( ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ15)
550
ﻟﺴﻠﺴﻠﺔPACF ﻭﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ ﺍﻟﺠﺯﺌﻲACF ﺩﺍﻟﺘﻲ ﺍﻟﺘﺭﺍﺒﻁ ﺍﻟﺫﺍﺘﻲ: 17 -15 ﺸﻜل ﻟﻠﺴﻠﺴﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺍﻟﺴﺎﺒﻘﺔARIMA(0,2,2) ﻟﻨﻤﻭﺫﺝ ﺒﻭﻜﺱ ﺠﻴﻨﻜﻨﺯresiduals ﺍﻷﺨﻁﺎﺀ
Error for WELLS from ARIMA, MOD_16 CON 1.0
.5
0.0
ACF
-.5
Confidence Limits
Coefficient
-1.0 1
3 2
5 4
7 6
9 8
11 10
13 12
15 14
16
Lag Number
Error for WELLS from ARIMA, MOD_16 CON 1.0
.5
Partial ACF
0.0
-.5
Confidence Limits
-1.0
Coefficient 1
3 2
5 4
Lag Number
7 6
9 8
11 10
13 12
15 14
16
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
551
ﻭﺒﺭﺴﻡ ﺸﻜل ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺍﻷﺼﻠﻴﺔ ﻤﻊ ﺴﻠﺴﻠﺔ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺘﻭﻗﻌﺔ ﺒﺎﻻﻋﺘﻤﺎﺩ ﻋﻠﻰ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﺫﻱ ﺘﻡ ﺘﻘﺩﻴﺭﻩ ) ARIMA(0,2,2ﺒﺎﻹﻀﺎﻓﺔ ﺍﻟﺘﻲ ﺴﻠﺴﻠﺘﻲ ﺤﺩﻱ ﺍﻟﺜﻘﺔ
ﺍﻟﺴﻔﻠﻴﺔ ﻭﺍﻟﻌﻠﻭﻴﺔ ﻟﻠﻘﻴﻡ ﺍﻟﻤﺘﻭﻗﻌﺔ ﻨﺤﺼل ﻋﻠﻰ ﺸﻜل 18-15ﺍﻟﺘﺎﻟﻲ ،ﻭﻫﺫﻩ ﺍﻟﺸﻜل
ﻴﻭﻀﺢ ﻤﺩﻯ ﺍﻟﺩﻗﺔ ﻓﻲ ﺘﻤﺜﻴل ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﻤﻘﺘﺭﺡ ﻟﺒﻴﺎﻨﺎﺕ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺍﻟﺘﻲ ﺘﻤﺜل
ﺃﻋﺩﺍﺩ ﺤﻘﻭل ﺍﻟﻨﻔﻁ ﺍﻟﻌﺎﻤﻠﺔ ﻓﻲ 50ﺸﻬﺭﹰﺍ ﻓﻲ ﺍﻟﻭﻻﻴﺎﺕ ﺍﻟﻤﺘﺤﺩﺓ ﺍﻷﻤﺭﻴﻜﻴﺔ ﺤﻴﺙ ﺘﺘﻘﺎﺭﺏ ﻜل ﻤﻥ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺍﻷﺼﻠﻴﺔ ﻤﻊ ﺴﻠﺴﻠﺔ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺘﻭﻗﻌﺔ.
ﺸﻜل : 18-15ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ ﺍﻷﺼﻠﻴﺔ ﻭﺴﻠﺴﻠﺔ ﺍﻟﻘﻴﻡ ﺍﻟﻤﺘﻭﻗﻌﺔ ﻭﺍﻟﻤﺘﻨﺒﺄ ﺒﻬﺎ ﺒﺎﻻﻋﺘﻤﺎﺩ ﻋﻠﻰ ﺍﻟﻨﻤﻭﺫﺝ ) ARIMA(0,2,2ﻭﻜﺫﻟﻙ ﺴﻠﺴﻠﺘﻲ ﺤﺩﻱ ﺍﻟﺜﻘﺔ ﻟﻠﻘﻴﻡ ﺍﻟﻤﺘﻭﻗﻌﺔ ﻭﺍﻟﻤﺘﻨﺒﺄ ﺒﻬﺎ 1200000
1000000
800000
WELLS Fit for WELLS from A RIMA, MOD_16 CON
600000
95% LCL for WELLS fr om ARIMA, MOD_16 CON
400000
95% UCL for WELLS fr om ARIMA, MOD_16 CON
200000 55 58
49 52
43 46
37 40
31 34
25 28
19 22
13 16
7 10
1 4
Sequence number
) (15ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺍﻟﺘﻨﺒﺅ ﺍﻹﺤﺼﺎﺌﻲ
552
.7 .15ﺗﺤﻠﻴﻞ اﻟﺴﻼﺳﻞ اﻟﺰﻣﻨﻴﺔ اﻟﻤﻮﺳﻤﻴﺔ : Analysis of Seasonal Time Series : ﻭﻓﻲ ﺍﻟﻨﻬﺎﻴﺔ ،ﻓﺈﻥ ﻓﻜﺭﺓ ﺘﻭﻓﻴﻕ ﻨﻤﻭﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ) ARIMA(p,I,qﻴﻤﻜﻥ ﺒﺒﺴﺎﻁﺔ ﺘﻁﻭﻴﺭﻫﺎ ﻟﺘﺸﻤل ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﺍﻟﻤﻭﺴﻤﻴﺔ Seasonal time Series
ﺒﺎﺴﺘﺨﺩﺍﻡ ﻨﻤﺎﺫﺝ ﺒﻭﻜﺱ-ﺠﻴﻨﻜﻨﺯ ARIMAﺍﻟﻤﻭﺴﻤﻴﺔ ﻭﺫﻟﻙ ﺒﺎﺴﺘﺨﺩﺍﻡ ﺍﻟﻨﻤﻭﺫﺝ
ARIMA(p,I,q)(P,I,Q)sﺤﻴﺙ Pﻭ Qﻫﻤﺎ ﻋﺎﻤﻠﻲ ﺍﻻﻨﺤﺩﺍﺭ ﺍﻟﺫﺍﺘﻲ ﻭﺍﻟﻤﺘﻭﺴﻁﺎﺕ ﺍﻟﻤﺘﺤﺭﻜﺔ ﺍﻟﻤﻭﺴﻤﻴﻴﻥ ﻭ sﻫﻲ ﻋﺩﺩ ﺍﻟﻔﺼﻭل ﺍﻟﻤﻭﺴﻤﻴﺔ.
ﻭﻴﺴﺘﻁﻴﻊ ﺍﻟﻘﺎﺭﺉ ﺘﻁﺒﻴﻕ ﻨﻤﺎﺫﺝ ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﺍﻟﻤﻭﺴﻤﻴﺔ Seasonal time
Seriesﺍﻟﺴﺎﺒﻘﺔ ﺍﻟﺫﻜﺭ ﻋﻠﻰ ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﺍﻟﻤﻭﺴﻤﻴﺔ ﺒﻌﺩ ﺍﺴﺘﻴﻌﺎﺏ ﺃﺩﻭﺍﺕ ﺍﻟﺘﻌﺭﻑ ﻋﻠﻰ ﺍﻟﻨﻤﺎﺫﺝ ﺍﻟﻤﺨﺘﻠﻔﺔ ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﻭﺘﻨﻔﻴﺫ ﺍﻷﻭﺍﻤﺭ ﺍﻟﺨﺎﺼﺔ ﺒﻨﻅﺎﻡ SPSSﺒﺩﻗﺔ، ﻭﻟﻜﻨﻨﺎ ﻟﻥ ﻨﺨﻭﺽ ﻓﻲ ﺘﻔﺎﺼﻴل ﺇﻀﺎﻓﻴﺔ ﻫﻨﺎ.
ﻜﻤﺎ ﺃﻨﻪ ﻴﻤﻜﻥ ﺘﺤﻠﻴل ﺍﻟﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ ﺍﻟﻤﻭﺴﻤﻴﺔ ﺒﻁﺭﻴﻘﺔ ﺍﻟﺘﺤﻠﻴل ﺇﻟﻰ
ﺍﻟﻤﺭﻜﺒﺎﺕ ﺍﻷﺴﺎﺴﻴﺔ ) Decompositionﻟﻠﺘﻔﺎﺼﻴل ﺃﻨﻅﺭ ﻋﻜﺎﺸﺔ ،(2001 ،ﻭﻫﺫﻩ
ﻁﺭﻴﻘﺔ ﺴﻬﻠﺔ ﺍﻟﺘﻁﺒﻴﻕ ﺭﻏﻡ ﺃﻨﻬﺎ ﺘﻔﺘﻘﺭ ﺇﻟﻰ ﺍﻟﺩﻗﺔ ﺃﺤﻴﺎﻨﹰﺎ ﻭﺍﻟﻤﺭﻭﻨﺔ ﻓﻲ ﺍﺨﺘﻴﺎﺭ ﺍﻟﻨﻤﻭﺫﺝ ﺍﻟﻤﻨﺎﺴﺏ ﺒﺎﻹﻀﺎﻓﺔ ﺇﻟﻰ ﻋﺩﻡ ﺇﻤﻜﺎﻨﻴﺔ ﺍﺴﺘﺨﺩﺍﻡ ﺃﺩﻭﺍﺕ ﺍﻻﺴﺘﻨﺘﺎﺝ ﺍﻹﺤﺼﺎﺌﻲ ﻋﻠﻰ ﻨﺘﺎﺌﺠﻬﺎ ،ﻭﻟﻜﻲ ﻴﺘﻡ ﺍﺴﺘﺨﺩﺍﻡ ﻫﺫﻩ ﺍﻟﻁﺭﻴﻘﺔ ﻻ ﺒﺩ ﻓﻲ ﺍﻟﺒﺩﺍﻴﺔ ﻤﻥ ﺘﻌﺭﻴﻑ ﻤﺘﻐﻴﺭ ﺘﺎﺭﻴﺦ
Date variableﻋﻥ ﻁﺭﻴﻕ ﺍﺨﺘﻴﺎﺭ ﺃﻤﺭ ﺘﻌﺭﻴﻑ ﺘﺎﺭﻴﺦ Define Datesﻤﻥ ﻗﺎﺌﻤﺔ ﺍﻟﺒﻴﺎﻨﺎﺕ Dataﻭﺍﺨﺘﻴﺎﺭ ﺸﻜل ﺍﻟﺘﺎﺭﻴﺦ Date Formatﺍﻟﺫﻱ ﻴﻨﺎﺴﺏ ﺍﻟﺴﻠﺴﻠﺔ ﺍﻟﺯﻤﻨﻴﺔ،
ﺜﻡ ﺒﻌﺩ ﺫﻟﻙ ﺍﺨﺘﺎﺭ ﺃﻤﺭ ﺍﻟﺘﺤﻠﻴل ﺇﻟﻰ ﺍﻟﻤﺭﻜﺒﺎﺕ ﺍﻷﺴﺎﺴﻴﺔ Decompositionﻤﻥ ﺍﻟﻘﺎﺌﻤﺔ ﺍﻟﻔﺭﻋﻴﺔ ﻟﻠﺴﻼﺴل ﺍﻟﺯﻤﻨﻴﺔ Time Seriesﻓﻲ ﻗﺎﺌﻤﺔ ﺍﻟﺘﺤﻠﻴل ﺍﻹﺤﺼﺎﺌﻲ
ﺍﻟﺭﺌﻴﺴﻴﺔ ﻟﺘﻔﺘﺢ ﺍﻟﻨﺎﻓﺫﺓ ﺍﻟﺨﺎﺼﺔ ﺒﺎﻟﺘﺤﻠﻴل ﺇﻟﻰ ﺍﻟﻤﺭﻜﺒﺎﺕ ﺍﻷﺴﺎﺴﻴﺔ .Decomposition ﻭﺭﻏﻡ ﺃﻥ ﻫﺫﻩ ﺍﻟﻁﺭﻴﻘﺔ ﻤﺒﺎﺸﺭﺓ ﻭﺘﻌﻁﻲ ﻨﺘﺎﺌﺞ ﻴﺴﻬل ﺍﻟﺘﻌﺎﻤل ﻤﻌﻬﺎ ﻭﺘﻔﺴﻴﺭﻫﺎ
ﺇﻻ ﺃﻨﻨﺎ ﻟﻥ ﻨﺨﻭﺽ ﻓﻲ ﺘﻔﺎﺼﻴل ﺃﻜﺜﺭ ﻤﻥ ﺫﻟﻙ ﻭﺫﻟﻙ ﻟﻸﺴﺒﺎﺏ ﺍﻟﺴﺎﺒﻘﺔ ﺍﻟﺫﻜﺭ.