Mining Attribute-structure Correlated Patterns in Large Attributed Graphs

《Mining Attribute-structure Correlated Patterns in Large Attributed Graphs》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、MiningAttribute-structureCorrelatedPatternsinLargeAttributedGraphsArleiSilvaWagnerMeiraJr.MohammedJ.ZakiUniversidadeFederaldeUniversidadeFederaldeRensselaerPolytechnicMinasGeraisMinasGeraisInstituteBeloHorizonte,BrasilBeloHorizonte,BrasilTroy,NYarlei@dcc.

2、ufmg.brmeira@dcc.ufmg.brzaki@cs.rpi.eduABSTRACTtheorganizationofverticesintodensesubgraphs.Forin-stance,weaimtoaddressquestionssuchas:HowdoesaInthiswork,westudythecorrelationbetweenattributesetsparticularsetofinterestsinducecommunitiesinasocialandtheoccur

3、renceofdensesubgraphsinlargeattributednetwork?Whatarethecommunitiesthatemergearoundgraphs,ataskwecallstructuralcorrelationpatternmin-suchinterests?Suchquestionsarerelatedtoimportantso-ing.Astructuralcorrelationpatternisadensesubgraphcialphenomenasuchashom

4、ophily[11]andinfluence[2].inducedbyaparticularattributeset.ExistingmethodsareAlthoughseveraldefinitionsofdensesubgraphshavebeennotabletoextractrelevantknowledgeregardinghowvertexproposedintheliterature,mostofthemdonottakevertexattributesinteractwithdensesub

5、graphs.Structuralcorre-attributesintoconsideration.Furthermore,suchdefinitionslationpatternminingcombinesaspectsoffrequentitemsetdonotprovideanyknowledgeregardinghowdifferentsetsandquasi-cliqueminingproblems.Weproposestatisticalofattributesinducedensesubgra

6、phs.significancemeasuresthatcomparethestructuralcorrela-Thisworkstudiesthecorrelationbetweenvertexattributestionofattributesetsagainsttheirexpectedvaluesusingnullanddensesubgraphs,ataskwecallstructuralcorrelationmodels.Moreover,weevaluatetheinterestingness

7、ofstruc-patternmining.Thestructuralcorrelationofanattributeturalcorrelationpatternsintermsofsizeanddensity.Ansetistheprobabilityofavertextobememberofadenseefficientalgorithmthatcombinessearchandpruningstrate-subgraphinitsinducedgraph.Moreover,astructuralcor

8、-giesintheidentificationofthemostrelevantstructuralcor-relationpatternisadensesubgraphinducedbyaparticularrelationpatternsispresented.Weapplyourmethodforattributeset.Figure1illustratesadatasetforstructuraltheanalysis