支持向量机分类器在医疗诊断中的应用研究

2001-12-19202200321006-9348200302-0069-02200030UCIbenchmarkTp181A!StatisticaILearningTheory1Sup-portVectorMachines233#!Ilx1y1xlylRIX+1-1PRIPx=1x2xNxyx=sgnwPx+6yiPxiw+61-EiEiwminwE=12ww+cli=1Ei1s.tyiPxiw+61-EiEi0i=1lcLw6Eav=12ww+cli=1Ei-li=1aiyiwPxi+6-1+Ei-li=1viEi2aiviai0vi0i=1lLw=0L6=0LE=03aiyi=0w=aiyiPxic-ai-vi=0442maxWa=-12li=1aiayiyPxiPx+li=1ai5PxiPxKxixKxixMercermaxWa=-12li-1aiayiyKXix+li-1ai696s.tli=l!iyi=00!ici=ll!iminl2xTHx+fTx7s.taTx=00xcx=aHi=yiyKxixii=llfT=-l-l-l-l!T=yly2yl-lylTKTaiyiKxxi+6=08fxw6=sgnaiyiKxxi+()69!!Kxxi3lKxxi=xxi+lii2Kxxi=KIx-xiI3Kxxi=Sxxi+cSSigmoid#CieveriandCiinicFoundationV.A.MedivcaiCenterRobertDetranoUCI4Cieveriand303755Vaiue0l234l3AgeSexCpTrestbpsChoiFbsRestecgThaiachExangOidpeakSiopeCaThaipresenceVaiuel234absenceVaiue0l3#$xi=xi-minximaxxi-minxil0xi0lpresence+lAbsence-lACAXC0lX+l-l#!Kxxi=exp-#Ix-xiI2${)2llIx-xiI=KI=lxIi-xI2#$##Cieveiand200Pentium350MHZl28MBMatiab6.0$%SVM90.81l.2783.51l.92l20!l00%NTgroeth577CLASSIT578.9C4574.5SVM83.5l2ll-lll2563071.M.1988.2.M.1999.3.J.1995331-5.4.J.199617918-23.1973.5-1942.9-1973.12-ComputerSimulationofStrategicEnvironmentalAssessmentZHANGYan1SHANGJin-cheng2YUXiang-yi21.ChangchunInstituteofGeographytheChineseAcademyofSciencesChangchunJiiin130021China2.TheInstituteofEnvironmentaiSciencesNortheastNormaiUniversityChangchunJiiin130024ChinaABSTRACTOnthebaseoftheobjectofsustainabiedeveiopmentthepaperstudiesthemethodofSEAprovidesthemodeiofassessmenttestsandverifiesthemodeibytheexampieofChangchun.Theresuitmakesciearthecomponentsofsystemsub-systemisnoniinearthereiationsofthemiscoherent.KEYWORDSSystemDynamicsStrategicEnvironmentai!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!SustainabieDeveiopment70141VapnikVN.TheNatureofsta-tisticaiLearningTheoryM.NewYorkSpringer-Veriag1995.2VapnikVN.AnoverviewofstatisticaiLearningTheoryJ.IEEETrans.NeuraiNetwork1999105988-999.3VapnikVN.TheNatureofstatisticaiLearningTheoryM.NewYorkSpringer-Veriag1999.4L.Precheit.Probeni.AsetofneurainetworkbenchmarkprobiemandbenchmarkruiesJ.FakuiitatfurinformatikUniv.KarisruheGer-manyTech.Rep.21/94sepy1994.5PeterD.Turney.Cost-SensitiveCiassificationEmpiricaiEvaiuationofaHybridGeneticDecisionTreeInductionAigorithmJ.JournaiofArtificiaiInteiiigenceResearch21995369-409.1972-1936-36ApplicationofSupportVectorMachinesandLeastSguaresSupportVectorMachinestoHeartDiseasediagnosesYANWei-wuSHAOHui-heDepartmentofAutomationShanghaiJiaotongUniversityShanghai200030ChinaABSTRACTMedicaIdiagnosiscarryoutbydatafrommuItipIepathoIogicaIexaminations.ForthedataarecharacterizedbyindividuaIspeci-ficityinherentnoisyitisdifficuIttoaccurateIydiagnose.SupportVectorMachinesSVMisanoveIpowerfuIIearningmethoddeveIopedonStatisticaILearningTheoryandofmanyspeciaIadvantages.SVMnonIinearcIassifiersaIgorithmisdiscussedinthepaper.UsingradiaIbasisfunctionskerneIsSVMnonIinearcIassifierisempIoyedtoheartdiseasediagnosesbasedonUCIbenchmarkdataset.ComparingotherresuIthighaccuracyrateisobtainedintheprediction.AppIicationofSVMtodiseasediagnosesindicatesSVMpotentiaIappIicationinmedicaI.KEYWORDSSupportvectormachineCIassifiers!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!Diagnoses119610090%51JYangWLuandAWaibeI.Skin-coIormodeIingandadaptationR.TechnicaIReportCMU-CS-97-146CMUPittsburghPA1997.2HARowIeySBaIujaT.Kanade.NeuraInetwork-basedfacedetec-tionJ.PatternAnaIysisandMachineInteIIigenceIEEETransPA-MI.VoIume20Issue1Jan.1998Pages23-38.3RobertoBruneIIiTomasoPoggio.FacerecognitionfeaturesversustempIatesJ.IEEETransPAMI.199315101042-1052.4LeungTKBurIMCPeronaP.FindingfacesincIutteredscenesus-ingrandomIabeIedgraphmatchingJ.FifthIntI.Conf.onComp.Vi-sionCambridgeMAJune1995.5ReisfeIdDYeshurunY.PreprocessingoffaceimagesdetectionoffeaturesandposenormaIizationJ.IEEETrans.ComputerVisionandImageUnderstanding1998371413-420.6.J.199927812-15.7.J.2001205375-380.8RRameshRKasturiandBSchunckMachineVisionM.pp31-51McgrawHiIINewYork1995.1976-1973-1939-FaceDetectionandFacialFeatureExtractionBasedonTheGeneralizedSymmetryTransformDUPingZHANgYan-kunLIUChong-gingInstituteofImageProcessingandPatternRecognitionShanghaiJiaoTongUniv.Shanghai200030ChinaABSTRACTInthispaperanapproachtofacedetectionbasedonskincoIormodeIandgeneraIizedsymmetrytransformisintroduced.WiththestabiIityofhumanskincoIorsdistributioninthechromaticspacetheskincoIorareasaresegmented.SincethetwoeyesonthefacearehighIysymmetricaIthegeneraIizedsymmetrytransformisthenempIoyedtoIocatetheeyesontheskincoIorregion.FinaIIytheaverageeyeimagesareusedastempIatetomatchtheeyeIocation.TheresuItoftheexperimentprovedtheefficiencyoftheaIgorithm.KEYWORDSFacedetectionFaciaIfeatureCoIormodeISymmetrytransformTempIatematch46支持向量机分类器在医疗诊断中的应用研究作者：阎威武，邵惠鹤作者单位：上海交通大学自动化研究所,上海,200030刊名：计算机仿真英文刊名：COMPUTERSIMULATION年，卷(期)：2003,20(2)被引用次数：12次参考文献(5条)1.PeterDTurneyCost-SensitiveClassification:EmpiricalEvaluationofaHybridGeneticDecisionTreeInductionAlgorithm19952.LPrecheltProbenl.Asetofneuralnetworkbenchmarkproblemandbenchmarkrules19943.VapnikVNTheNatureofStatisticalLearningTheory19994.VapnikVNAnoverviewofstatisticalLearningTheory[外