资源描述：
《evolutionary algorithms for data mining》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。
1、Chapter20EVOLUTIONARYALGORITHMSFORDATAMININGAlexA.FreitasUniversiryofKent,UKAbstractEvolutionaryAlgorithms(EAs)arestochasticsearchalgorithmsinspiredbytheprocessofDarwinianevolution.ThemotivationforapplyingEAstoDataMin-ingisthattheyarerobust,adaptivesearchtech
2、niquesthatperformaglobalsearchinthesolutionspace.ThischapterreviewsmainlytwokindsofEAs,viz.GeneticAlgorithms(GAS)andGeneticProgramming(GP),anddiscusseshowEAscanbeappliedtoseveralDataMiningtasks,namely:discoveryofclassificationrules,clustering,attributeselecti
3、onandattributeconstruction.ItalsodiscussesthebasicideaofMulti-ObjectiveEAs,basedontheconceptofParetodominance,whichalsohasapplicationsinDataMining.Keywords:geneticalgorithm,geneticprogramming,classification,clustering,attributese-lection,attributeconstruction
4、,multi-objectiveoptimization1.IntroductionTheparadigmofEvolutionaryAlgorithms(EAs)consistsofstochasticsearchalgorithmsinspiredbytheprocessofDarwinianevolution(BackandWeigend,1998;EibenandSmith,2003).EAsworkwithapopulationofin-dividuals,eachofthemacandidatesol
5、utiontoagivenproblem,that"evolve"towardsbetterandbettersolutionstothatproblem.Itshouldbenotedthatthisisaverygenericsearchparadigm.EAscanbeusedtosolvemanydifferentkindsofproblems,bycarefullyspecifyingwhatkindofcandidatesolutionanindividualrepresentsandhowthequ
6、alityofthatsolutionisevaluated(bya"fitness"function).Inessence,themotivationforapplyingEAstoDataMiningisthatEAsarerobust,adaptivesearchmethodsthatperformaglobalsearchinthespaceofcandidatesolutions.Incontrast,severalmoreconventionalDataMiningmeth-odsperformalo
7、cal,greedysearchinthespaceofcandidatesolutions.Asa436DATAMININGANDKNOWLEDGEDISCOVERYHANDBOOKresultoftheirglobalsearch,EAstendtocopebetterwithattributeinteractionsthangreedyDataMiningmethods(Freitas,2002A;Freitas,2002C;DharandStein,1997;PapagelisandKalles,2001
8、;Freitas,2001).Hence,intuitivelyEAscandiscoverinterestingknowledgethatwouldbemissedbyagreedymethod.Theremainderofthischapterisorganizedasfollows.Section2presentsanoverviewofEAs,inordertom
