搜索
36420094ComputerScienceVol.36No.4Apr.2009:2008205229(60773130),(07JC630034)(1967-),,,,E2mail:pengyan@mail.cnu.edu.cn;(1953-),,,;(1972-),,,;(1935-),,,1112(100037)1(100083)2,,(CBR,CasebasedReasoning)(SVM,SupportVectorMachine)(ANN,ArtificialNeuralNetwork),,,,,,ResearchofMachineLearning2basedRiskPredictionPENGYan1WANGWan2sen1WANGXu2ren1TUXu2yan2(CollegeofInformationEngineering,CapitalNormalUniversity,Beijing100037,China)1(CollegeofInformationEngineering,USTB,Beijing100083,China)2AbstractPredictiontechnologyistopredictthedeveloptendencyandfuturestateofaevenwithscientificmethodandmeans.Inordertoremedytheinsufficientoftraditionalapproach,variousmachinelearningtechnicalhasbeenappliedinforecastresearch.Thispaperdiscussedtheuseofmachinelearningintheriskpredictionfield,suchasSupportsVectorMachine,CasebasedReasoningaswellasArtificialNeuralNetwork.Atthesametime,basedonourownwork,theap2plicationofmachinelearning2basedpredictionmodelusedincreditriskpredictionandfoundationprojectbidwasdis2cussedindetail.KeywordsMachinelearning,Prediction,CBR,SVM,ANN1,,200,20,,,,,,(MDA),,;,,,,,,,,:(CBR,CasebasedReasoning)(SVM,SupportVectorMachine)(ANN,ArtificialNeuralNetwork),,,2,,,4[1]502,,()(),,,2.1,,,,CBR,,,AarmodtPlazaCBR[2],4:(1);(2);(3);(4),,,,,,HongkyuIngoo[3,4]4()3()3:,Jha,HuiFoo[5]CBRANN,CBR,CBR,Jarillo,SucciW.Pedrycz[6]CBR,ANNs,RI,3:ANN,RIKimHan[7]SOMCBR,CBR,,SOM/CBR,EssamAhmed[8],CBR,,ANNs,2.2SupportVectorMachines(SVMs)Vap2nik[9],,(OSH),SVM,SVM[10],2000-20021888,944944,:,,(PCA)(0.5),SVM,,,SVM,BPN,MDALogit,,SVM(),SVMBPN,SVM,SVM2.3(ANN,ArtificialNeuralNetwork),()(),,,ANN,ANN,,ANN,n,u,ANN:,,,,,BP602ANN,,,,TamKiang[11],DuttaShekhar[12]3BP,,,()33.1,25,,1,SOM2PNNSOM,,()PNNSOM,PNN,,2,,,,(),3.2,(AHP)(SOM),,3:AHPSOM3;,SOM,SOM3AHP2SOFM15,,[0,1],15SOFM,3:(10.8)(0.80.6)(0.60)44,,,,,(),,,,,,[1],,.[J].,2003,17(2):45248[2]AarmodtA,PlazaE.Case2basedreasoning:foundationalissues,methodicalvariationsandsystemapproaches.AICommunica2tions,1994,7(1):39259[3]JoHongkyu,HanIngoo,LeeHoonyoung.Bankruptcypredic2tionusingcase2basedreasoning,neuralnetworks,anddiscriminantanalysis.ExpertSystemswithApplications,1997,13(2):972108[4]JoHongkyu,HanIngoo.Integrationofcase2basedforecasting,neuralnetwork,anddiscriminantanalysisforbankruptcypredic2tion.ExpertSystemswithApplications,1996,11(4):4152422[5]JhaG,HuiSC,FooS.Ahybridcase2basedreasoningandneu2ralnetworkapproachtoonlinefaultdiagnosisProc.3rdInter2nationalICSCSymposiaonIntelligentIndustrialAutomation(IIA99)andSoftComputing(SOCO99).Genoa,Italy,1999:3762381(210)702,,RSVM(breast),,,RSVMFSVM(2),,12v2112v2a,3,12v2112v2a,i2ris,auto,vehicle33,(RSVM)SVMFSVM;machinepolynomialRBFglassRBF,RSVMSVMFSVM(3),1-3,,,(RSVM)SVMFSVM2polynomial()PolynomialDataSVMFSVM(100%)(90%)(80%)(70%)breast0.9474(26)0.9649(25)RSVM:0.9649(24)RSVM:0.9474(25)RSVM:0.9474(25)RSVM:0.9474(24)pima0.7865(302)0.7760(292)RSVM:0.7865(300)RSVM:0.7865(304)RSVM:0.7917(304)RSVM:0.7917(305)heart0.7432(89)0.7432(87)RSVM:0.7432(89)RSVM:0.7838(88)RSVM:0.7432(87)RSVM:0.7432(90)bupa0.6353(198)0.6118(175)RSVM:0.6353(198)RSVM:0.6353(197)RSVM:0.6353(196)RSVM:0.6353(196),,,RSVMSVMFSVM,[1]CortesC,VapnikV.SupportVectorNetworks.Machinelear2ning,1995,20(3):2732297[2]VapnikVN.EstimationofDependenciesBasedonEmpiricalDa2ta.Berlin:Springer2Verlag,1982[3]HsuChihwei,LinChihjen.Acomparisonofmethodsformulti2classsupportvectormachines.IEEETransactionsonNeuralNetworks,2002,13(2):4152425[4]HuangHP,LiuYH.Fuzzysupportvectormachinesforpatternrecognitionanddatamining.Int’lJournalofFuzzySystems,2002,4(3):8262835[5]LeeKi2Young,KimDae2Won.Possibilisticsupportvectorma2chines.PatternRecognition,2005,38(8):132521327[6]BoLiefeng,JiaoLicheng,WangLing.WorkingSetSelectionU2singFunctionalGainforLS2SVM.IEEETransactionsonNeuralNetworks,2007,18(5):15412154[7]PawlakZ.Roughset.InternationalJournalofInformationandComputerScience,1982,11(5):3412356(207)[6]JarilloG,SucciG,PedryczW,etal.AnalysisofSoftwareEngi2neeringDataUsingComputationalIntelligenceTechniquesProceedingsofthe7thInternationalConferenceonObject2O2rientedInformationSystems.Calgary,AB,2001[7]KimKyung2Sup,HanIngoo.Thecluster2indexingmethodforcase2basedreasoningusingself2organizingmapsandlearningvectorquantizationforbondratingcases.ExpertSystemswithApplications,2001,21(3):1472156[8]ChangchienaSW,LinbMing2Chin.Designandimplementationofacase2basedreasoningsystemformarketingplans,2005,28(1):43253[9]VapnikVN.TheNatureofStatisticalLearningTheory.Sprin2ger,1995[10]MinJH,LeeYoung2Chan.Bankruptcypredictionusingsupportvectormachinewithoptimalchoiceofkernelfunctionparame2ters.ExpertSystemswithApplications,2005,28:6032614[11]TamKY,KiangMY.ManagerialApplicationsofNeuralNet2works:TheCaseofBankFailurePredictions.ManagementS2cience,1992,38:9262947[12]DuttaS,ShekharS.BondRating:ANon2ConservativeApplica2tionofNeuralNetworksProceedingsoftheIEEEInternatio2nalConferenceonNeuralNetworks.SanDiego,NewYork:IEEEPress,1989:4432450012