ImprovedFeatureSelectionAlgorithmBasedon
SVMandCorrelation
Zong-XiaXie,Qing-HuaHu,andDa-RenYu
HarbinInstituteofTechnology,Harbin,P.R.China
xiezongxia@hcms.hit.edu.cn
Abstract.Asafeatureselectionmethod,supportvectormachines-recursivefeatureelimination(SVM-RFE)canremoveirrelevancefea-turesbutdon’ttakeredundantfeaturesintoconsideration.Inthispa-per,itisshownwhythismethodcan’tremoveredundantfeaturesandanimprovedtechniqueispresented.Correlationcoe cientisintroducedtomeasuretheredundancyintheselectedsubsetwithSVM-RFE.Thefeatureswhichhaveagreatcorrelationcoe cientwithsomeimportantfeatureareremoved.ExperimentalresultsshowthatthereactuallyareseveralstronglyredundantfeaturesintheselectedsubsetsbySVM-RFE.Thecoe cientsarehighto0.99.Theproposedmethodcannotonlyre-ducethenumberoffeatures,butalsokeeptheclassi cationaccuracy.1Introduction
Inrecentyears,datahasbecomeincreasinglylargernotonlyinnumberofin-stancesbutalsoinnumberoffeaturesinmanyapplications,suchasgeneselec-tionfrommicro-arraydataandtextautomaticcategorization,etc.,wherethenumberoffeaturesintherawdatarangesfromhundredstotensofthousands.Highdimensionalitybringsgreatdi cultyinunderstandingthedata,construct-ingclassi ersystemsforpatternrecognitionandmachinelearning.Therefore,Featureselectionhasattractedmuchattentionfrompatternrecognitionandmachinelearningsociety[1,2,3,4].TheaimofFSisto ndasubsetoforigi-nalfeaturesfromagivendatasetthroughremovingirrelevantand/orredundantfeatures[5].FScanimprovethepredictionperformanceofthepredictions,pro-videfasterandmorecost-e ectivepredictorsandgiveabetterinsightoftheunderlyingprocessthatgeneratedthedata[3,4].
Traditionalstatisticalmethodsforclusteringandclassi cationhavebeenex-tensivelyusedforfeatureselection.SupportVectorMachines(SVMs),proposedin1990sbyVapnik,havealreadybeenknownasane cienttoolfordiscover-inginformativeattributes[6].SVMsminimizetheupperboundoftrueerrorbymaximizingmarginwithwhichtheyseparatethedataintodistinctclasses.Intheframeworkofstatisticallearningtheory,theupperboundoferrorsofaclassi erdependsonitsVCdimensionandthenumberofthetrainingdata.FeaturesubsetselectionmayleadtoreducetheVCdimensionofthetrainedclassi er,andthenimprovestheclassi cationperformance.Alotofempiricalanalysisshowedintheliteraturessupportsthisargument.
J.Wangetal.(Eds.):ISNN2006,LNCS3971,pp.1373–1380,2006.
cSpringer-VerlagBerlinHeidelberg2006
Feature selection algorithm for text classification based on improved mutual information_专业资料。Ju~lfHriIsteoTcnly(eei)o.8,o321 oraoab tu ehoo nnitfgNwSrs...
Network Security Incidents Frequency Prediction Based on Improved Genetic Algorithm and LSSVM_专业资料。哼{eaouan 遏之 hmnts iCmii nco Netor ...
algorithm based on SVM is presented , and its key techniques in the process of text organizing2text expression , feature selection , SVM algorithm , etc....
Improved PSO algorithm based on chaos theory and its application to design... Feature selection algo... 15人阅读 5页 2.002014 Baidu 使用百度...
【Abstract】An improved feature selection algorithm based on gradient is ...【Key words】Support Vector Machine(SVM); feature selection; kernel function...
An improved branch and bound algorithm for feature selection_专业资料。...Based on this observation, Yu and Yuan (1993) proposed the BB+ algorithm...
AN IMPROVED FAST BLIND DECONVOLUTION ALGORITHM BASED ON DECORRELATION AND BLOCK MATRIX_专业资料。维普资讯 http://www.cqvip.com VI5N.o_ 052 JURLO LCONCCNA...
3. Algorithm description In recent years, various feature selection algorithms...The improved gini index only used the number of documents which contain the...
【Abstract】An improved feature selection algorithm based on gradient is ...【Key words】Support Vector Machine(SVM); feature selection; kernel function...
classification performances are kept or improved. ... feature selection, sample selection, SVM 1 ...algorithm requires solving a quadratic programming (...
热门文档