一種基于變長編碼的遺傳K均值算法研究.pdf

1、洳：鏘蛐Z摯碩1學位論文⑧論文題口二拙墓變量纏型曲塑佳K塑值簋這盟塑作旨姓名————撞，∞＆衛(wèi)指導教師學科(專業(yè))所在學院提交乃塒———直班』蛙L——————一——j刪至攫——————一丑‘鹱扭量熟鹽堂瞳———墊Ⅲ=￡絲———————一浙江大學碩士學位論文AbstractAbstractThecomplexrelationshipsamongdatacoverpatternsthatpeopleCallhardlydiscoverOne

2、ofthosepatternsiStoclassifyobjectsthedatarepresentsRecentlydataminingtheoryhasbeendevelopedforawiderangeofclusteringanalysisalgorithmtoanalyzedatatouncoverthepatternTheyalIhaveadvantagesanddisadvantagesandseveralofthemha

3、vebeenputintopracticeForexampleK—McansalgorithmiSasimpleandeffectivemethod，butithasmanydisadvantagesToavoidthese，researchershaveraisedtheimprovedKmeansalgorithmAsweknowgeneticalgorithmfGAlhasbeenprovedeffectivetosolvepro

4、blemslikecombinationoptimizeandextremesofafunctionItsmechanismiSfitforimprovingKMcansclusteringaswellAlthoughtherearemanyGAmethodsbeingusedtoimproveinitiaicentersselectionofclustering。ortolearnthebestValueofknoneoftllemc

5、allperfectlycombinebotIlofthemtogetherBeingenlightenedanewGAmethodbasedonvariablelengthencodingiSputforwardforpurpose。andreceivedagoodeffectThegeneticKMcansalgorithm(OKA)basedonvariable—lengthencodingnotonlyoptimizedthei

6、nitialcenters。anddynamicallylearnedthevalueofkbutalSecognizedSomeisolatedpointsSoitreducedtheimpactofisolatedpointstoKMeansA11thisdependsonagoodmechanismoflearningKandanexcellentfitnessfunctionInordertoIearnktheindividua

7、lofbestfitnessWastakenastheexample；everynewindividualgeneratedshouldgetacloselengthtoitAccordingly’userdoesn’tneedtospeeifyKanymoreForabetterunderstanding，thepaperemphasizetheexplainingofbasictheoryofdatamining(DM)and0AB