全部 标题 作者
关键词 摘要

OALib Journal期刊
ISSN: 2333-9721
费用:99美元

查看量下载量

相关文章

更多...

一种过滤式多标签特征选择算法

DOI: 10.13232/j.cnki.jnju.2015.04.010, PP. 723-730

Keywords: 多标签,特征选择,过滤式,互信息,标签相关性

Full-Text   Cite this paper   Add to My Lib

Abstract:

多标签数据的过滤式特征选择依靠特征评价对特征选择,快速有效得到候选特征。但现有算法多将标签集合简单化,将其视作独立标签加以研究,忽视了多标签集合内部相互关系。近年来,由于mrmr算法在单标签数据领域方面简单、快速、高效的特征选择能力,成为过滤式特征选择算法的流行算法之一。提出一种基于mrmr(max-relevancemin-redundancy)过滤式多标签特征选择算法(ml-mrmr),直接通过对特征进行权重计算,得到特征与多标签集合的相互关系,以获得更好的候选特征子集。同时,算法的特征评价过程中不仅考虑了特征间以及特征与多标签的相互影响,更考虑到多标签内部可能存在的相互关系,将标签相关性加入特征评价当中,提出了可适应多标签数据的度量标准。最后,在真实多标签数据集上的实验结果表明:所提算法能够对数据大幅降维并稳定有效地提高降维后数据的分类效果。

References

[1]  .hanjw.datamining:conceptsandtechniques.the2ndedition.北京:机械工业出版社,2001.
[2]  .spolaorn,chermanea,monardea,etal.acomparisonofmulti-labelfeatureselectionmethodsusingtheproblemtransformationapproach.electronicnotesintheoreticalcomputerscience,2013,292:135?151.
[3]  .dendamrongvits,vateekulp,kubatm.irrelevantattributesandimbalancedclassesinmulti-labeltext-categorizationdomains.intelligentdataanalysis,2011,15(6):843-859.
[4]  .pereirarb,plastinoa,zadroznyb,etal.categorizingfeatureselectionmethodsformulti-labelproblems.knowledge-basedsystems,2014.
[5]  .李思男,李宁,李战怀.多标签数据挖掘技术研究综述.计算机科学,2013,40(4):14-21.
[6]  .mulan:ajavalibraryformulti-labellearning.http://mulan.sourceforge.net.
[7]  .tsoumakasg,katakesi,vlahavassi.dataminingandknowledgediscoveryhandbook.springer,berlin,2010:667-685.
[8]  .zhangml,zhouzh.areviewonmulti-labellearningalgorithms.ieeetransactiononknowledgeanddataengineering,2014,26(8):1819-1837.
[9]  .周国静,李云.基于最小最大策略的集成特征选择.南京大学学报(自然科学),2014,50(4):457-465.
[10]  .spolaorn,monardmc,leehd.asystematicreviewtoidentifyfeatureselectionpublicationsinmulti-labeleddata.icmctechnicalreport,2012,374:31.
[11]  .王露,龚光红.基于relieff_mrmr特征降维算法的多特征遥感图像分类.中国体视学与图像分析,2014,19(3):250-257.
[12]  .guyoni,elisseeffa.anintroductiontofeatureextraction.in:featureextraction,foundationsandapplications.springer,2006:1?24.
[13]  .chenwz,yanj,zhangby,etal.documenttransformationformulti-labelfeatureselectionintextcategorization.in:the7thieeeinternationalconferenceondatamining,2007,451?456.
[14]  .zhangml,zhouzh.ml-knn:alazylearningapproachtomulti-labellearning.patternrecognition,2007,40(7):2038?2048.
[15]  .trohidisk,tsoumakasg,kallirisg,etal.multi-labelclassificationofmusicintoemotions.in:the9thinternationalconferenceonmusicinformationretrieval,2008,325-330.
[16]  .leej,kimdw.featureselectionformulti-labelclassificationusingmulti-variatemutualinformation.patternrecognitionletters,2013,34(3):349?357.
[17]  .shaoh,ligz,liugp,etal.symptomselectionformulti-labeldataofinquirydiagnosisintraditionalchinesemedicine.sciencechinainformationsciences,2012,54(1):1-13.
[18]  .guqq,lizh,hanjw.correlatedmulti-labelfeatureselection.in:the20thacminternationalconferenceoninformationandknowledgemanagement,2011,1087-1096.
[19]  .lastrag,luaceso,quevedojr,etal.graphicalfeatureselectionformulti-labelclassificationtasks.in:the10thinternationalconferenceonadvancesinintelligentdataanalysisx,2011,246?257.
[20]  .kongxn,yups.gmlc:amulti-labelfeatureselectionframeworkforgraphclassification.knowledgeinformationsystems,2012,31(2):281?305.
[21]  .pupoogr,morellc,sotosv.relieff-ml:anextensionofreliefalgorithmtomulti-labellearning.progressinpatternrecognition,imageanalysis,computervision,andapplications,springer,2013,528?535.
[22]  .readj,pfahringerb,holmesg.multi-labelclassificationusingensemblesofprunedsets.icdm’08,2008,995?1000.

Full-Text

Contact Us

[email protected]

QQ:3279437679

WhatsApp +8615387084133