智能控制智能控制 (9).pdfVIP

4IEEETRANSACTIONSONEVOLUTIONARYCOMPUTATION,VOL.18,NO.1,FEBRUARY2014

ASurveyofMultiobjectiveEvolutionary

AlgorithmsforDataMining:PartI

AnirbanMukhopadhyay,SeniorMember,IEEE,UjjwalMaulik,SeniorMember,IEEE,

SanghamitraBandyopadhyay,SeniorMember,IEEE,andCarlosArtemioCoelloCoello,Fellow,IEEE

Abstract—Theaimofanydataminingtechniqueistobuildalgorithmsisoftennotfeasible.Applicationsofevolutionary

anefficientpredictiveordescriptivemodelofalargeamountofalgorithms,withtheirinherentparallelarchitecture,havebeen

data.Applicationsofevolutionaryalgorithmshavebeenfoundfoundtobepotentiallyusefulforautomaticprocessingoflarge

tobeparticularlyusefulforautomaticprocessingoflarge

quantitiesofrawnoisydataforoptimalparametersettingandamountsofrawnoisydataforoptimalparametersettingand

todiscoversignificantandmeaningfulinformation.Manyreal-todiscoversignificantandmeaningfulinformation[2],[3].

lifedataminingproblemsinvolvemultipleconflictingmeasuresTraditionally,evolutionaryalgorithms(EAs)[4]wereused

ofperformance,orobjectives,whichneedtobeoptimizedtosolvesingleobjectiveproblems.However,manyreal-life

simultaneously.Underthiscontext,multiobjectiveevolutionaryproblemshavemultipleconflictingperformancemeasuresor

algorithmsaregraduallyfindingmoreandmoreapplications

inthedomainofdataminingsincethebeginningofthelastobjectives,whichmustbeoptimizedsimultaneouslytoachieve

decade.Inthistwo-partpaper,wehavemadeacomprehen