基于強化學習的績優(yōu)股票預測系統(tǒng)研究.pdf

1、國內(nèi)圖書分類號：TPl81國際圖書分類號：62L3工學碩士學位論文Y867325基于強化學習的績優(yōu)股票預測系統(tǒng)研究碩士研究生：導師：申請學位級別：學科、專業(yè)：所在單位：授予學位單位：楊櫻葉德謙教授工學碩上計算機應用技術信息科學與工程學院燕山大學鲞當盔蘭三蘭堡主蘭壘堡蘭AbstractAsestablishingnonlinearmodeusingartificialneuralnetworkhasbeenwidelyusedtoecon

2、omyforecastingforrecentlyyears，thispaperwilldosomeresearchonthetopicUsingBPneuralnetworksOnstockforecastingsystemisanewresearchfieldandhasagreatevolvementButtlliSmethodisnotgoodatstabilityanddefinitionSoitmakethesystem’S

3、selfstudyabilitynotstrong，whenthedataquantumbecomebigger，theforecastingresukbecomeweakerrefereacevalueInthispaperweprovidestheoptimizationonareinforcementlearningalgorithmbasedonneuralnetworkUsingthismethodwecanimproveth

4、esystem’Sstabilityanddefinitionandimprovethegeneralizationabilityoflearningsystem(1)Choosepropersocketdatatooperate，assurethedatahaveadefinitereferencevalue(2)Choosethepropermethodtopretreatmentthedata，assurethedatahaveg

5、oodastringencyandfasterconstringencyrateduringtheBPneuralnetworkstrainingprocess，thesystemperformancegaineddefiniteensure(3)Adopttheappropriatealgorithmbasedonneuralnetworkensemble，assuretheforecastingsystem’sstabilizati

6、onconquertheBPneuralnetworksystem’SinstabilitySothattheforecastingresulthavehigherreferencevalueandtheforecastingstockpricecalvehasbetterstability似)Analysisthecharacteristicoftimeserialadopttherightreinforcementlearninga

7、lgorithmtooptimizetheresultoftheneuralnetwork’SoutputchoosetheexcellentparameterofthereinforcementlearningsystemAllalgorithmsaboveareimplementedinMATLABandutilizingVCandMATLABrealizationthesystemconfirmexcellentparameter