基于智能計(jì)算的一元時間序列分析與預(yù)測.pdf

1、西安電子科技大學(xué)碩士學(xué)位論文基于智能計(jì)算的一元時間序列分析與預(yù)測姓名：楊承勇申請學(xué)位級別：碩士專業(yè)：計(jì)算機(jī)應(yīng)用技術(shù)指導(dǎo)教師：劉芳20040101AbstractAbstractUnvariedTimeSeriesanalysisandforecastingisanimportantportionofcurrentsignalprocessandeconomics.Itscoreproblemistoestablishthedynamic

2、modeltotheinsidecontactoftheseriesandthenmakesuseofthemodelproceedtheforecasting.Itprimarilyincludestwocontents:constructingpredictorandtheresearchofpredictoralgorithm.Sointhispaperourworksmainlyconcentrateon:thefurtherd

3、evelopmentofthecurrentpredictorandtheimprovementofpredictoralgorithm..ImmunityMixAlgorithmbasedonthecontinuousdiferentialfunctionisproposedinthispaperanditsastringencyisproved.Fromherewegettheaccurateestimatormethodofthe

4、AutoregressiveMovingAveragemodelcoefficient.Thealgorithmissimulatedwiththetestfunctionofoptimizationandactualtimeseriesobtainingthegoodperformance...ForthegeneralseasontimeseriesaccordingtothemodelofSeasonAutoregressiveI

5、ntegratedMovingAverageTheconceptofhorizontalandlengthwaystrendaregaveandanewseasontimeseriesmodelisbroughtforward.Andthentheprocessofmodelingissimplifiedconsumedly.Totheestimateproblemofakindoftimeseriesthemodelperforman

6、ceisgoodForgeneralnonlineartimeseries(notseasontimeseries)onthefoundationofPipelinedRecurrentNeuralNetworkBFGS(BroydenFletcherGoldfardShannon)isintroducedinsoastudyalgorithmbasedonBFGSisputforward.Thealgorithmspeedsupnet

7、workstudytraincomparedwiththeRealtimeRecurrentLeamingalgorithmconsumedlyandincreasestheaccuracyofprediction.Thesimulationandcontrastexperimentsresultareprovidedindetailinthispaper.Someexperienceregulationsaresummedup.Key