一個二階段覆蓋聚類算法及其應用.pdf

1、分類號UDC密級學號200301263延邊大學碩士掌位論文一個二階段覆蓋聚類算法及其應用研究生姓名培養(yǎng)單位指導教師姓名、職稱學科專業(yè)研究方向洪義成理學院數(shù)學系姜今錫教授基礎數(shù)學數(shù)理統(tǒng)計論文提交日期二零零六年六月二日AbstractClusteranalysisisakindofmultivariatestatisticalmethodofsampleclassificationanddataanalysiswithouttraining

2、samplesetWhenweutilizethismethodtoclassifythegivendata，thoughwedon’tknowthetypeofadoptedsampleitwillclassifythesamplesautomaticallyaccordingtotheirsimilarityThemainobjectiveinclusteringanalysisistodiscovernaturalgrouping

3、softhedata，SOthatthedatawithineachgroupaxerelativelysimilar(ie，theypossesslargelythesamecharacteristics)audtheobservationsindifferentgroupsarerelativelydissimilarAgoodresultofclustering，ontheonehand，wecallgrasPtheeachgro

4、upwithgivendata’Scharacteristicsaccordingtotheirinherentnature，SOitcomesuptotheobjectofafewunderlyingandisusefulforfurtherdiscussionandresearchtothegivenproblemInthisthesis，weproposetwostepcoveringclusteraigofithmbytheid

5、eaofcover—ingalgorithmMeanwhile，intheprocessofanalyzingsomeclusteringproblems，whichincludessomeinter—relevantbetweencharactervariable，WeusePrincipalComponentAnalysistoovercomebadinfluenceonstabilityofclusteringsomevariab

6、leswhicharehighlycorrelatedWetrytopromotethespeedofclusteringalgorithm，andguaranteethevalidoftheclusteringresult，andthroughsomeanalysisofexamplestoillustrateandexaminethefeasibilityandvahdatioaoftheputtingforwardmethodKe