支持向量机及其在铁路工程中的应用研究

《支持向量机及其在铁路工程中的应用研究》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、西南交通大学硕士学位论文支持向量机及其在铁路工程中的应用研究姓名：蒋学申请学位级别：硕士专业：道路与铁道工程指导教师：李远富20081001西南交通大学硕士研究生学位论文第1I页AbstractSupportVectorMachine(SVM)basedontheSLT(statisticallearningtheory)isanewmachinelearningmethod，whichwasdevelopedbyVapnikandhisteamin1995；itembodiesthetheoryofstructureriskminimization(S

2、RM)andCansolvetheproblemcharacterizedbynonlinear,highdimension，smallsampleandlocalmi‘ni‘mi。zi‘ngperfectly．SVMhasbecomethehotspotinthefieldofmachinelearningbecauseofitsexcellentlearningperformance，SOitwasappliedsuccessfullyinmanyengineeringfields．ThisthesisappliedSVMtotherailwayen

3、gineefing，becauseithasbetterlearningfeatureandfutureapplyvaluewhichishopedtosolvemuchprobleminthedatamining(DM)．ThefollowingworkWasconductedinthisthesis：．1．Expatiatedandcomparedthecommonmethodsinthepassengerpredictionandinvestmentprediction．Detailedlydeducedthetraininganddecision

4、·-makingprocessofSVMfromlinearSVMtoNon··linearSVMandsumthetrainingalgorithm．2．Summarizedthetheoryofneuralnetwork，constructedthepassengervolumepredictionmodelbasedonBPandrailwayinvestmentpredictionmodelbasedonRBFbyprogrammingonMATLABlanguage．3．Constructedtheregressionmodelonthebas

5、eoftheoryofSVM，andthenappliedthismodelonthepredictionofcityrailwaypassengervolumebyLIBSVM．ThencomparedtheresultsbetweenSVMandBP,resultindicatethatSVMismoreprecisethanNNinthesituationofsmallsample．4．Predictionoftheinvestmentisthehotspotofrailwayproject，constructedthesupportvectorr

6、egressionmodelbyLIBSVM，andapplieditontheTBMrailwaytunnelcostprediction，resultshowthatSVMhaveabetterperformancethanNNinthesituationofhighdimension．111eresultsshowthatSVMmethodiSbetterthanneuralnetwork．ItiSbelievedthatUsingSVMtheorytOsolveregressionproblemisamethodwithpromisingpros

7、pect．Atlast，apersonalpreviewoffurthertasksintheresearchrealmsofneuralnetworkandsupportvectormachineiSpresented．西南交通大学硕士研究生学位论文第1II页-_●_●■●■_●I●____●■●■■■■●■●■_●■●■■●■●mn-ml●■l●●●■●_■●■●■l_■●■_●■_■_●_■■●_●●l●■●■____●Keywords：supportvectormachine；neuralnetwork；regressionmodel；railw

8、aypassengervolume；investmentprediction西南