基于遗传算法的模糊聚类挖掘方法应用研究

《基于遗传算法的模糊聚类挖掘方法应用研究》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、ABSTRACTWithinformationtechnologyandveryfastspeed，informationprocessingdatabasetechnologydevelopingathasbecomeaindispensabletoolforpeopletoacquireusefulmessage．Dataminingisagenericknowledgediscoverytechnology，itisaprocessoffindingmodelandtherelationshipofthedatainalargeamountofdatabyanalyti

2、caltools．Clusteringanalyticalisanimportantcomponentofdataminingtechnology．Dataclusteringminingtechnologyisanemergingareawhichinvolvesvariousareas·FCM(Fuzzyc．means)algorithm，asakindofunsupervisedlearningmethods，itisaresearchhotspotconcerningaboutdataclusteringanalyticaltechnology．FCMisoneofi

3、mportantalgorithmindataclusteringminingmethods．ithasthecharacteristicsassimple，fastconvergenceandstronglocalsearchingpower，etc．However，FCMissensitivetoinitializationandtendstoresultinlocalminimuminiterations．GeneticAlgorithmisarandomsearchingglobaloptimizationalgorithm．Itisacomputationalmod

4、elofthehumanevolution，withimplicitparallelismandcapacityofusingeffectiveglobalinformation．ThecombinationofFCMalgorithmandgeneticalgorithmwillgetahybridalgorithmwhichbenefitstosolveclusteringproblemandmaketremendousimprovementinalgorithmperformance，thehybridalgorithmhasgoodglobalandlocalsear

5、chcapability．ThispaperpresentsahybridFuzzyC-meansalgorithm(IG·FCM)basedonimprovedgeneticalgorithm．Thealgorithmuseglobalsearchabilityofgeneticalgorithmtooptimizetheinitialclustercentersofclusteringalgorithm，andthencarryouttheFCMalgorithmbaseonlocaloptimization·IG．FCMisaHeuristicclusteringalg

6、orithm，itorderlychangesclusteringclassnumber，thenautomaticallydeterminetheoptimalnumberofclusteringclassandtheoptimalclusteringbaseonEvaluationofclusteringvalidityfunction．Becauseofthetraditionalgeneticalgorithmhasshortcomingslikeslowconvergence，poorstabilityandlowaccuracyrate·Thispaperadop

7、tstheoptimumpreservationstrategyintheselectionoperationtomaintaintheoptimumindividualintheprocessofgenetic，andIllthencopytheselectedindividual，thentheoptimumindividualentersthenextgenerationdirectlywithoutparticipatingincrossoverandmutatIonoperation．Thec