A Novel Visual analytics Approach for Clustering Large-Scale Social Data

《A Novel Visual analytics Approach for Clustering Large-Scale Social Data》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、2013IEEEInternationalConferenceonBigDataANovelVisualanalyticsApproachforClusteringLarge-ScaleSocialDataZhangyeWangJuanxiaZhouWeiChenStateKeyLabofCAD&CGStateKeyLabofCAD&CGStateKeyLabofCAD&CGZhejiangUniversityZhejiangUniversityZhejiangUniversityHangzhou,ChinaHangzhou,ChinaHangzhou,Chinazywan

2、g@cad.zju.edu.cnxjzhou@zjucadcg.zju.edu.cnchenwei@cad.zju.edu.cnChangChenJiyuanLiaoRossMaciejewskiCollegeofSoftwareEngineeringStateKeyLabofCAD&CGSchoolofCIDSEUniversityofScienceandTechnologyofChinaZhejiangUniversityArizonaStateUniversityHefei,ChinaHangzhou,ChinaUSAlawangel@163.comjyliao@zj

3、ucadcg.zju.edu.cnrmacieje@asu.eduAbstract—Socialdatareferstodataindividualscreatethatistypeofdataasusertagdata.knowinglyandvoluntarilysharedbythemandisanexcitingGiventheubiquityofsuchdata,methodsforexploringavenueintogaininginsightintointerpersonalbehaviorsandandanalyzingsocialdatalinkshav

4、ebecomeacriticalinteraction.However,suchdataislarge,heterogeneousandresearchtopic.Byanalyzingusertagsanduserbehavioroftenincomplete,propertiesthatmaketheanalysisofsuchdataextremelychallenging.Onecommonmethodofexploringdata,itisexpectedthatinsightsintosocietalpatternscansuchdataisthroughclu

5、steranalysis,whichcanenableanalystsbegleanedandusedtoinformgovernmentalpolicyandtofindgroupsofrelatedusers,behaviorsandinteractions.Thisbusinessdecisionsaswellasinfluencehumanbehavior[2].paperpresentsanovelvisualanalysisapproachfordetectingForexample,previouswork[3]exploresclusteringusersclu

6、sterswithinlarge-scalesocialnetworksbyutilizingadivide-intogroupstodeveloptargetedmarketstrategybasedonanalyze-recombineschemethatsequentiallyperformsdatapartitioning,subsetclusteringandresultrecombinationwithindemographicsandbehavioralpatterns.However,ourabil-anintegratedvisualinterface.A

7、casestudyonamicroblogitytoextractknowledgefromtheselargeheterogeneousmessagingdata(with4.8millionsusers)isusedtodemonstratedatasourcesisstilllimitedduetothelatent,subtleandthefeasibilityofthisapproachandcomparisonsarealsounpredictablerelationshipsthatmaybehidd