KMEANS聚类算法研究及应用

《KMEANS聚类算法研究及应用》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、武汉理工大学硕士学位论文K-means聚类算法研究及应用姓名：张建辉申请学位级别：硕士专业：系统工程指导教师：王学锋20070401武汉理I：大学硕士学何论文ABSTRACTClusteringisamajorfieldindataminingwhichalsoisanimportantmethodofdatapartitionorgrouping．Clusteringnowhasbeenappliedintovariouswaysincommerce，marketanalysis，biology,Webclass

2、ificationandsoon．Andclusteringalgorithmsincludespartitioning,hierarchical，density-based，grid·based，model-basedalgorithmandfuzzyclustering．K-meansalgorithmisoneoftheessentialclusteringalgorithms．Itisal【indofclusteringalgorithmbasedonpartitioningmethod．Thethesis

3、isplannedtoimprovethealgorithmbasedontheresearch，whileontheapplicationaspect，thethesistakethealgorithmintothecustomersegmentationuse．CustomersegmentationistheessentialelementfortheenterprisetotakeoutCRM．Inthefirstpartofthepaper，itshowsthemainelaborationobjecto

4、fstudybackgroundmaterialaswellasthegoalthisarticlemustachieve，andshowstheresearchthementalityandtheoverallcontent．Thesecondpartmainlyintroducesthebasicknowledgeofclusteringandmethodsforclusteringanalysis，onthebasisofthedifferentalgorithmsanalysis，gettingtheadv

5、antagesanddisadvantagesthroughCompareofdifferentalgorithms．Thethirdpartintroducetheapplicationofthealgorithm，thethesisappliestheclusteringtechnologyintheuseofthecustomersegmentation，FirstbuildingthecustomervaluesystemthroughAHP；quantifycustomervalue；thendivide

6、customerintodifferentclassificationsusingclusteringtechnology,socantakeouttheefficientCRMaccordingtothedeferentcustomerclassification．Currentlytherearesomeevaluatingsystemsofcustomervalue，butDoneofthemcanbeputintopracticeefficiently,Thethesisbuiltanevaluatings

7、ystemofcustomervaluewhichisinlinewiththedevelopmentoftheenterprise，usingthemethodofdatamining,basedonthepracticalsituationoftheenterpriseandthroughseriesofpracticalevaluatingindexofcustomervalue，TheevaluatingsystemCallbeusedtoquantifycustomervalue，segmentcusto

8、merandbuilddecisionsupportingsystemofthecustomervaluemanagement，Thefourthpartisthecurepart，mainlyanalysisthetypicalalgorithm一一K-means，thethesisproposedtwoalgorithmstoimprovethek-me