ImbalancedSVMClassificationMethodBasedonIncrementalLearning

doi:10.3969/j.issn.1006-2475.2018.07.004

Abstract

Abstract: Thispaperpresentsanimbalancedsupportvectormachine(SVM)basedonincrementallearning,namelyISVM_IL,tosolvetheimbalancedclassificationproblemthatthetraditionalSVMclassificationmethodcannotsolve.Firstly,thismethodextractssomesamplesfromthemajoritynegativeclass,andtheinitialclassifiercanbeobtainedbytrainingSVMonthesesamplesandminoritypositiveclasssample.Then,accordingtotherelationshipbetweentheclassifierandothernegativesamples,thenearestsampletotheclassifierisselectedasanincrementalsampletojointhetrainingsettoparticipateintheSVMtraining.Therefore,thenegativeclasssizeoftheactualtrainingisreducedandtheperformanceofimbalancedclassificationisimproved.TheexperimentresultdemonstratesthattheproposedISVM_ILmethodcanimprovetheclassificationperformanceofimportantminorityclasssampleofimbalancedclassification.

Key words: supportvectormachine, imbalancedclassification, incrementallearning, ISVM_ILalgorithm

CLC Number:

TP18

CUILi-na1,GUOHu-sheng2. ImbalancedSVMClassificationMethodBasedonIncrementalLearning[J]. Computer and Modernization, doi: 10.3969/j.issn.1006-2475.2018.07.004.

References

［1］叶志飞,文益民,吕宝粮.不平衡分类问题研究综述［J］.智能系统学报,2009,4(2):148-156.
［2］SunYanmin,WongA,KamelSM,etal.Classificationofimbalanceddata:Areview［J］.InternationalJournalofPatternRecognitionandArtificialIntelligence,2009,23(4):687-719.
［3］周志华.机器学习［M］.北京:清华大学出版社,2016.
［4］JamesM,MichaelC,BradB,etal.BigData:TheNextFrontierforInnovation,Competition,andProductivity［R］.McKinseyGlobalInstitute,2011.
［5］郭虎升.支持向量机建模及应用研究［D］.太原:山西大学，2014.
［6］CharteF,RiveraA,JesusM,etal.MLSMOTE:Approachingimbalancedmultilabellearningthroughsyntheticinstancegeneration［J］.KnowledgeBasedSystems,2015,89:385-397.
［7］KrawczykB,GalarM,HerreraF.Evolutionaryundersamplingboostingforimbalancedclassificationofbreastcancermalignancy［J］.AppliedSoftComputing,2016,38(C):714-726.
［8］TangYuchun,ZhangYanqing,ChawlaNV,etal.SVMsmodelingforhighlyimbalancedclassification［J］.IEEETransactionsonSystems,ManandCybernetics,PartB(Cybernetics),2009,39(1):281-288.
［9］AggarwalSK,GoelA,SinghVP.StageanddischargeforecastingbySVMandANNtechniques［J］.WaterResourcesManagement,2012,26(13):3705-3724.
［10］PangShaoning,KasabovN.R-SVMT:DiscoveringtheknowledgeofassociationruleoverSVMclassificationtrees［C］//ProceedingsoftheInternationalJointConferenceonNeuralNetworks(IJCNN).2008:2486-2493.
［11］〖ZK(＃〗ZhangZhiwang,GaoGuangxia,ShiYong.Creditriskevaluationusingmulti-criteriaoptimizationclassifierwithkernel,fuzzificationandpenaltyfactors［J］.EuropeanJournalofOperationalResearch,2014,237(1):335-348.
［12］陶卿,梁万路,孔康,等.加权边缘损失函数的代价敏感支持向量机［J］.模式识别与人工智能，2011,24(6):763-768.
［13］LiuYang,AnAijun,HuangXiangji.BoostingpredictionaccuracyonimbalanceddatasetswithSVMensembles［C］//Proceedingsofthe10thPacific-AsiaConferenceonKnowledgeDiscoveryandDataMining(PAKDD).2006:107-118.
［14］AkbaniR,KwekS,JapkowiczN.Applyingsupportvectormachinestoimbalanceddatasets［C］//ProceedingsofEuropeanConferenceonMachineLearning(ECML).2004:39-50.
［15］郭虎升,王文剑.处理非平衡数据的粒度SVM学习算法［J］.计算机工程,2010,36(2):181-183.
［16］UCI.UCIMachineLearningRepository［EB/OL］.http://archive.ics.uci.edu/ml,2009-10-16.