聚类分析文献英文翻译Word格式.doc

聚类分析文献英文翻译Word格式.doc

《聚类分析文献英文翻译Word格式.doc》由会员分享，可在线阅读，更多相关《聚类分析文献英文翻译Word格式.doc（14页珍藏版）》请在冰豆网上搜索。

译文名称：

数据挖掘—聚类分析

专业：

自动化

姓名：

****

班级学号：

****

指导教师：

******

译文出处：

Datamining：

IanH.Witten,EibeFrank著

二○一○年四月二十六日

Clustering

5.1INTRODUCTION

Clusteringissimilartoclassificationinthatdataaregrouped.However,unlikeclassification,thegroupsarenotpredefined.Instead,thegroupingisaccomplishedbyfindingsimilaritiesbetweendataaccordingtocharacteristicsfoundintheactualdata.Thegroupsarecalledclusters.Someauthorsviewclusteringasaspecialtypeofclassification.Inthistext,however,wefollowamoreconventionalviewinthatthetwoaredifferent.Manydefinitionsforclustershavebeenproposed:

lSetoflikeelements.Elementsfromdifferentclustersarenotalike.

lThedistancebetweenpointsinaclusterislessthanthedistancebetweenapointintheclusterandanypointoutsideit.

Atermsimilartoclusteringisdatabasesegmentation,whereliketuple（record）inadatabasearegroupedtogether.Thisisdonetopartitionorsegmentthedatabaseintocomponentsthatthengivetheuseramoregeneralviewofthedata.Inthiscasetext,wedonotdifferentiatebetweensegmentationandclustering.AsimpleexampleofclusteringisfoundinExample5.1.Thisexampleillustratesthefactthatthatdetermininghowtodotheclusteringisnotstraightforward.

AsillustratedinFigure5.1,agivensetofdatamaybeclusteredondifferentattributes.Hereagroupofhomesinageographicareaisshown.Thefirstfloortypeofclusteringisbasedonthelocationofthehome.Homesthataregeographicallyclosetoeachotherareclusteredtogether.Inthesecondclustering,homesaregroupedbasedonthesizeofthehouse.

Clusteringhasbeenusedinmanyapplicationdomains,includingbiology,medicine,anthropology,marketing,andeconomics.Clusteringapplicationsincludeplantandanimalclassification,diseaseclassification,imageprocessing,patternrecognition,anddocumentretrieval.Oneofthefirstdomainsinwhichclusteringwasusedwasbiologicaltaxonomy.RecentusesincludeexaminingWeblogdatatodetectusagepatterns.

Whenclusteringisappliedtoareal-worlddatabase,manyinterestingproblemsoccur:

lOutlierhandlingisdifficult.Heretheelementsdonotnaturallyfallintoanycluster.Theycanbeviewedassolitaryclusters.However,ifaclusteringalgorithmattemptstofindlargerclusters,theseoutlierswillbeforcedtobeplacedinsomecluster.Thisprocessmayresultinthecreationofpoorclustersbycombiningtwoexistingclustersandleavingtheoutlierinitsowncluster.

lDynamicdatainthedatabaseimpliesthatclustermembershipmaychangeovertime.

lInterpretingthesemanticmeaningofeachclustermaybedifficult.Withclassification,thelabelingoftheclassesisknownaheadoftime.However,withclustering,thismaynotbethecase.Thus,whentheclusteringprocessfinishescreatingasetofclusters,theexactmeaningofeachclustermaynotbeobvious.Hereiswhereadomainexpertisneededtoassignalabelorinterpretationforeachcluster.

lThereisnoonecorrectanswertoaclusteringproblem.Infact,manyanswersmaybefound.Theexactnumberofclustersrequiredisnoteasytodetermine.Again,adomainexpertmayberequired.Forexample,supposewehaveasetofdataaboutplantsthathavebeencollectedduringafieldtrip.Withoutanypriorknowledgeofplantclassification,ifweattempttodividethissetofdataintosimilargroupings,itwouldnotbeclearhowmanygroupsshouldbecreated.

lAnotherrelatedissueiswhatdatashouldbeusedofclustering.Unlikelearningduringaclassificationprocess,wherethereissomeaprioriknowledgeconcerningwhattheattributesofeachclassificationshouldbe,inclusteringwehavenosupervisedlearningtoaidtheprocess.Indeed,clusteringcanbeviewedassimilartounsupervisedlearning.

Wecanthensummarizesomebasicfeaturesofclustering（asopposedtoclassification）:

lThe（best）numberofclustersisnotknown.

lTheremaynotbeanyaprioriknowledgeconcerningtheclusters.

lClusterresultsaredynamic.

TheclusteringproblemisstatedasshowninDefinition5.1.Hereweassumethatthenumberofclusterstobecreatedisaninputvalue,k.Theactualcontent（andinterpretation）ofeachcluster,,,isdeterminedasaresultofthefunctiondefinition.Withoutlossofgenerality,wewillviewthattheresultofsolvingaclusteringproblemisthatasetofclustersiscreated:

K={}.

DEFINITION5.1.GivenadatabaseD={}oftuplesandanintegervaluek,theclusteringproblemistodefineamappingf:

whereeachisassignedtoonecluster,.Acluster,containspreciselythosetuplesmappedtoit;