资源描述:
《Interactive Local Clustering Operations for High Dimensional Data in Parallel Coordinates》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、InteractiveLocalClusteringOperationsforHighDimensionalDatainParallelCoordinatesPeihongGuoHeXiaoZuchaoWangXiaoruYuan∗KeyLaboratoryofMachinePerception(MinistryofEducation),andSchoolofEECSPekingUniversity,Beijing,P.R.China.ABSTRACTtechniqueshavebeenproposedtoclusterthedataandthusre
2、duceInthispaper,weproposeanapproachofclusteringdatainparal-visualcluttering[4,2,12,14,24];however,mosttechniquesarelelcoordinatesthroughinteractivelocaloperations.Differentfromeitherautomaticorsemi-automaticandusersareusuallyexcludedmanyothermethodsinwhichclusteringisgloballyapp
3、liedtothefromthecourseofvisualexplorationinthesensethattheyarenotwholedataset,ourinteractiveschemeallowsuserstodirectlyapplyactivelyengagedintheidentificationofclusters.Anothersubstan-attractiveandrepulsiveoperatorsatregionsofinterests,takingad-tialproblemisthatthesetechniquescan
4、notalwaysgeneratesat-vantagesofanelectricityinteractionmetaphor,forclutterreductionisfactoryclusters,sointhesecasesusersprobablywishtomakeandclusterdetection.Ourdesignenablesuserstointeractdirectlysomerefinementstotheclusteringresults.Althoughmostexistingwiththeparallelcoordinate
5、plotsandprovidesgreatflexibilityintechniquesprovidesomeadjustableparametersfortuningcluster-exploringandrevealingunderlyingpatterns.Withinstantfeedback,ingresults,theystilllacktheflexibilityofidentifyingclustersasourworkallowsuserstodynamicallyadjusttheclusteringparame-userswish.A
6、similarproblemexistsingraphvisualization.Astheterstoreachanoptimum.Wealsosupplytheuserwithagraphin-numberofnodesgraduallygoesup,edgeclutteringprogressivelydicatingthelogicalrelationshipbetweenclusters.Ourexperimentsinterfereswithusers’explorationandinterpretationofthegraph.showt
7、hatourschemeismoreefficientthantraditionalmethodsinWonget.al.suggestedintroducingEdgeLens[19]toimprovetheperformingvisualanalysistasks.visualqualityofcomplicatedgraphsbybendingtheedgesingraphswithamagnetmetaphor.However,themethodofEdgeLensmainlyKeywords:parallelcoordinates,high-d
8、imensionaldata,cluster-focusesongraphvisualizat