非平穩(wěn)時間序列的建模方法研究.pdf

1、武漢理工大學碩士學位論文非平穩(wěn)時間序列的建模方法研究姓名：林卉申請學位級別：碩士專業(yè)：應用數(shù)學指導教師：童恒慶20050501武漢理j一人學碩士學位論文AbstraetTimeseriesanalysisisoneofimportantbranchinprobabilityandstatisticsRecentlythetheoryandapplicationoftimeseriesisdevelopedquicklyAccording

2、totheeharacteristicofstatistics，timeseriesisdividedintotwoclasses0neisstationarytimeseries，theotherisnonstationarytimeseriesIndailytimeweusuallyobservetimeserieswhichisalmostnonstationaryespeciallyinthephenomenaofsociety

3、andeconomicThenonstationarytimeseriespresentsobvioilstendencyandperiodicityTheresearchoftllemodelingofnonstationarytimeseriesisveryimportantinpracticeThestudynarratesmainlyaboutthemethodofthemodelingofnonstationarytimese

4、riesFirstlysometraditionalresearchmethodsofthemodelingofnonstationarytimeseriesareintroducedinthestudyForexample，ARIMAmodel，seasonmodelX一11modeLgreymodelandsoonTheemphasisofthestudyistheresearchofthestatespacemodelingofn

5、onstationarytimeseries，itincludesthefivepoints：(i)Inmodelingphase，accordingtothemethodofKitagawaonthestatespacemodelingofnonstationarytimeseriestimeseriesisdecomposedintofouritemswhicharetrenditem，cycleitem，seasonitemand

6、randomitemThenthestudyestablishesthestatespacemodelingoffonritemsseparaflywhichconstitutethetotalstatespacemodelingofnonstationarytimeseries(ii)Intheestimateofstatevector，thestudyintroducestheacademicgistoftheKalmanfilte

7、ringandusesthearithmeticof也eKallllanfilteringandopdmalfixedintervalsmoothing(iii)ThestudyintroducestheimprovedEMalgorithm，anditprovestheconvergentproperty(iv)In血eestimateoftheparameterthestudyusesthearithmeticoftheKalman

8、filteringoptimalfixedintervalsmoothingandimprovedEMalgorithm(v)ThestudyusestheARIMAmodelandstatespacemodeltoforecastthetotalvalueofthethreeindustryinourcountryItcomparestheimpactofthetwomodelAndittestities也atthearithmeti

9、coftheKalmanfilteringandoptimalfixedintervalsmoothingisbetterthantheKalmanfilteringAndtheKalmanfilteringandthearithmeticoftheKalmanfilteringandoptimalfixedintervalsmoothingarcsteadyinarithmeticKeywords：nonStationarytimes