基于統(tǒng)計(jì)學(xué)習(xí)理論的證券市場(chǎng)預(yù)測(cè)與風(fēng)險(xiǎn)控制.pdf

1、上海交通大學(xué)博士后士學(xué)位論文基于統(tǒng)計(jì)學(xué)習(xí)理論的證券市場(chǎng)預(yù)測(cè)與風(fēng)險(xiǎn)控制姓名：楊一文申請(qǐng)學(xué)位級(jí)別：博士后士專業(yè)：企業(yè)管理指導(dǎo)教師：楊朝軍20041101上海交通太學(xué)安奏管理學(xué)院管理科學(xué)與rfit博士后流動(dòng)站出站報(bào)告說(shuō)明后者低估了風(fēng)險(xiǎn)。【關(guān)鍵詞】非線性時(shí)間序列分析統(tǒng)計(jì)學(xué)習(xí)理論支持向量機(jī)流動(dòng)性VaRAbstractOneofpurposesofdatadrivenmachinelearningistofindouttheregularities

2、，whichcan’tbediscoveredwithprincipleanalysis，toforecastthefuturedataWithexcellentabilityoffunctionapproximation，neuralnetworksarewidelyusedtodevelopthemapbetweenthepastandthefuturedatatocarryoutthepredictionsFirst，inchap

3、ter2，weanalyzetheshortcomingsoftheneuralnetworksbasedontheruleofempiricalriskminimization(ERM)，andintroducetheruleofstructuralriskminimization(SRM)toovercometheshortcomingsofERMSecond，vectorsupportmachine(SVM)，analgorith

4、mimplementingSRM，isintroducedFinallymulti—steppredictionsofthetrendofShanghaiSecurityCompositeIndexarcachievedwithacceptableaccuracyChapter3presentsamethodforpredictingthetrendofMockmarketbasedonsupportvectormachineandpr

5、ofitunityapproachbyWallStreetInvestmentExpertFirstthetrendpredictionisviewedasaproblemofpaRernrecognitionAndthenthepatternvectorsaleconstructedwiththehelpofprofitunityapproachFinallyonesteptrendprediction，thepaRemclassif