Graph mining: laws, tools, and case studies

209 Pages·2012·4.764 MB·English

by  Faloutsos, Christos;Chakrabarti, Deepayan

Series ISSN: 2151-0067 C
Series ISSSerNie: s2 I1S5S1N-0: 0261751-0067 C H C M
H A H MM
SYSNYSTNYHTNHETSHEISESI SSLI SLE LCETCEUTCRUTERUSER SOE SNO NON AKRA KRAB AKRA &&&CCCMMoMororgrgagnana & n & C & Cl aClyalpyaopyopoool olP luP ubPulbilbsilhsihsehrersesrs
DATA MINING AND KNOWLEDGE DISCOVERY B A B
DADTAA MTA IMNININGI NAGN DA NKDN KOWNOLEWDLGEED GDEI SDCIOSVCEORVYERY AR RT AR
T I • T
Series Editors: Jiawei Han, University of Illinois at Urbana-Champaign, I • FA FAL I • FA
SeriesS Eerdiietso Ersd: iJtioarwse: iJ Hiawanei, UHnainv,e Ursnitiyv oefr sIiltlyin oofi Is lalitn Uoirsb aatn Ua-rCbahnaam-pCahiagmn,paign, L O L
Lise Getoor, University of Maryland, Wei Wang, University of North Carolina, Chapel Hill, O U O
Lise GLeitsoeo Gr, eUtonoivr,e Ursnitiyv oefr sMitya royf lManadr, yWlaneid W, Waneig W, Uannigv,e Ursnitiyv oefr sNitoyr otfh NCoarrtohl iCnaa,r oClihnaap,e Cl Hhaiplle,l Hill, U T U Graph Mining
Johannes Gehrke, Cornell University, Robert Grossman, University of Illinois at Chicago T SO T GGrarpaphh M Minininingg
JohanJnoehs aGnnehesr kGe,e Chrokrne,e lCl oUrnneivlle Ursintiyv, eRrsoibtye, rRt oGbreorsts Gmraons, sUmnainv,e Ursnitiyv oefr sIiltlyin oofi Is lalitn Coihs iacat gCohicago S S
O S O
S S
Graph Mining
GraGprha pMhi Mnininging Laws, Tools, and Case Studies
Laws, Tools, and Case Studies LaLwasw, Ts,o Toloso, lasn, adn Cda Csea Sset uSdtuiedsies
LawLs, aTwoos,l sT, oaonlds, Canadse C Satused Sietsudies
Deepayan Chakrabarti, Facebook Inc.
DeepaDyeaenp Cayhaank rCahbaakrtria, bFaarcteib,o Foak cIenboc.ok Inc.
Christos Faloutsos, Carnegie Mellon University
ChrisCtohs rFisatloosu Ftsaolos,u Ctsaorsn,e Cgiae rMneeglileo nM Uelnloivne Ursintiyversity
G
What does the Web look like? How can we find patterns, communities, outliers, in a social network? Which G R G
What dWoehs atth ed oWese bth leo oWk elbik leo?o kH loikwe? c aHno wwe c fiann dw pea fittnedrn psa, tctoemrnms,u cnoimtimes,u onuittilieesr, so,u itnl iaer sso, ciina la n seotcwiaol rnke?t wWohrikc?h Which R A R
are the most central nodes in a network? These are the questions that motivate this work. Networks A P A
are thea mreo tsth ec emntorsat lc nenotdreasl inno ad ens eitnw ao rnke?t wTohreks?e Tarhee sthe ea rqeu tehstei oqnuse stthioatn sm tohtaivt amteo tthiviast we othrkis. Nwoetrwk.o Nrkestworks P H P
and graphs appear in many diverse settings, for example in social networks, computer-communication H M H
andn gertawapnohdrs kg sar pa(pipnehtasrr ua ispnipo mena adrn eiynt e dmcitviaeonrnys,e dt srieavtfetfriisnce gm sse,a tfntoiarng geexsm,a fmeonrp tel)ex, aipnmr osptoleceii nain-l p nsroeotctwieaoiln rn kiesnt, twceooramrckptsi,uo ctnoe rmn-ecptouwmtoemrr-kucson iminc ambtiiuoonlnoicgayt,ion M IN M
netdwoocrnukemst w(einnottr-rkutses (xiiotn nbt ridupesatireotcnitt eido gentr,ea tcprtahifosf niicn, tm rtaeafxnftia crg emetmraienevanagtl)e, ,m ppereronstote)ni, n-p-arpcocrtooetuineni-ntp grinorattepeirhnasc itinniot enfir nancaetntiwocinoa lrn kfersta wiunod br dkiosel tioengc tybi,ioonlo,gy, IN IN IN
documdeonctu-tmexetn bt-ipteaxrtt ibteip garratpithe sg irna ptehxst i rne ttreixetv rael,t rpieervsaol,n p-earcscoonu-natc gcoraupnht sg irna pfihnsa innc ifianl afrnacuiadl dfreateucdt idoent,ection, IN G IN
and others. G G
and otahnedrs .others. Deepayan Chakrabarti
In this work, first we list several surprising patterns that real graphs tend to follow. Then we give DeDepeaeypaany aCnh Cakhraakbraarbtai rti
In thisI wn othrkis, fiwrostr kw, efi rlisstt w seev leisrta ls esuverprarli ssiunrgp priastitnegr npsa tthteartn rse tahl agtr arepahl sg treanpdh st ot efnodll otow .f Tolhloewn. wTeh egniv ewe give
a detailed list of generators that try to mirror these patterns. Generators are important, because they
Christos Faloutsos
a detaiale dde ltiasitl eodf gliesnt eorfa gtoernse trhataot rtsr yth taot mtryir rtoor m thirersoer ptahtetseer npsa. tGteernnse.r aGtoernse arareto irms paroer tiamnpt,o bretacnaut,s bee tchaeuyse they
can help with “what if ” scenarios, extrapolations, and anonymization. Then we provide a list of ChCrihstroisst Fosa lFoaultosuotssos
can heclapn w hitehlp “ wwhitaht “iwf ”h sacte inf a”r sioces,n eaxritorasp, oexlattriaopnosl,a atinodn sa, naonndy manioznatyimoniz. aTthioenn. Tweh epnr owveid per ao vliidste oaf list of
powerful tools for graph analysis, and specifically spectral methods (Singular Value Decomposition
powerfpuol wtoeorflus lf otor oglrsa fpohr ganraaplyhs iasn, aalnydsi ssp, aencidfi scpalelcyi fsipceaclltyr aslp mecettrhalo mdse (tShiondgsu (lSarin Vgaulluaer DVaelcuoem Dpeocsoitmiopnosition
(SVD)), tensors, and case studies like the famous “pageRank” algorithm and the “HITS” algorithm
(SVD)()S, VteDns)o),r st,e annsodr sc,a asen dst cuadsiee ss tluikdei etsh lei kfea mthoeu fsa “mpaoguesR “panagke” Ralagnokri”t ahlmgo rainthdm th aen “dH tIhTe S“”H aIlgToSr”i tahlmgorithm
for ranking web search results. Finally, we conclude with a survey of tools and observations from
for ranfkoirn rga nwkeibn gs ewarecbh sreeasruclhts r. eFsiunltasl.l yF, iwnea lclyo, nwcelu cdoen wcliutdhe a w suitrhv eay s oufr vteoyo losf a tnodo los basnedrv oabtisoenrsv aftrioomns from
related fields like sociology, which provide complementary viewpoints.
relatedr efilealtdesd lfiikeeld sso lciikoel osgoyc,i owlhoigcyh, wprhoicvhid pe rcoovmidpel ecmomenptlaermy evnietawrpy oviinetws.points.
About SYNTHESIs
AbouAtb oSYuNtT SHYNESTIHs ESIs
This volume is a printed version of a work that appears in the Synthesis M
This voTluhmis ev oisl uam per iinst ead p vrienrtseiodn v oerfs aio wn oorfk at hwaot rakp tpheaatr sa pinp etahres Siny nththee sSiysnthesis M O M
Digital Library of Engineering and Computer Science. Synthesis Lectures
O R O
DigitalD Liigbirtaarl yL oibf rEanryg ionfe Eernignign aeenrdi nCgo amnpdu Cteorm Spciuetnecre S. Sciyenntche.e Ssiysn Ltheecstuisr eLsectures
provide concise, original presentations of important research and development R G R
providep croonvcidisee ,c oornicgiisnea, lo prirgeisneanlt aptrieosnesn toaft iimonpso orft aimntp roersteaanrct hr easnedar dchev aenlodp dmeveenltopment G A G
topics, published quickly, in digital and print formats. For more information A N A
topivciss,i tpt ouwpbwilciwssh,. pmeudob qrlgiusaihcnekcdlyl a,q iyunpi codkoilgyli.,t cianol madnigdi tparli annt dfo prrminatt sf.o Frmora mts.o Freo rin mfoorrme aintifoonrmation N & N
visit wwviwsi.tm wowrgwa.nmcolarygpaonocll.acyopmool.com ISBN: 978-1-60845-115-9 &C CL &C SYNTHESIS LECTURES ON
MoMroMgroagrnag &n a &Cn l &Ca lyCaplyoapoyoploo Plou Plbu lPbiuslbhislehirsehsresrsISBN: I9S7B8N-:1 -967088-4159--061001980504-00599-00100105-09 LAYP AYPO LAYP SDYDANSDTATAYTAHN AMTET AMSHI MINESIIN SNLIIINGNSEI GNLCA NGTAEUD NCARD NTKEU DKNSR K ONOEWONSN WOLOEWLNDELDGEEGD EGD EDI SDICSIOCSVOCEVOREVYREYRY
www.morganclaypool.com 9 781608 451159 O O O
www.wmwowr.gmanorcglaaynpcolaoylp.coooml.com 9 7816908784156101859451159 O L O
Jiawei Han, Lise Getoor, Wei Wang, Johannes Gehrke, Robert Grossman, Series Editors
L L
JiaweiJ iHawanei, LHiasen ,G Leistoe oGr, eWtoeoir ,W Waenig W, Joanhgan, Jnoehsa Gnneehsr kGe,e Rhrokbee, rRt oGbreorst sGmraons,s Smeraine,s SEedriietos rEsditors

Graph Mining
Laws,Tools,andCaseStudies
Synthesis Lectures on Data
Mining and Knowledge
Discovery
Editors
JiaweiHan,UIUC
LiseGetoor,UniversityofMaryland
WeiWang,UniversityofNorthCarolina,ChapelHill
JohannesGehrke,CornellUniversity
RobertGrossman,UniversityofChicago
SynthesisLecturesonDataMiningandKnowledgeDiscoveryiseditedbyJiaweiHan,Lise
Getoor,WeiWang,JohannesGehrke,andRobertGrossman.Theseriespublishes50-to150-page
publicationsontopicspertainingtodatamining,webmining,textmining,andknowledge
discovery,includingtutorialsandcasestudies.Thescopewilllargelyfollowthepurviewofpremier
computerscienceconferences,suchasKDD.Potentialtopicsinclude,butnotlimitedto,data
miningalgorithms,innovativedataminingapplications,dataminingsystems,miningtext,web
andsemi-structureddata,highperformanceandparallel/distributeddatamining,datamining
standards,dataminingandknowledgediscoveryframeworkandprocess,dataminingfoundations,
miningdatastreamsandsensordata,miningmulti-mediadata,miningsocialnetworksandgraph
data,miningspatialandtemporaldata,pre-processingandpost-processingindatamining,robust
andscalablestatisticalmethods,security,privacy,andadversarialdatamining,visualdatamining,
visualanalytics,anddatavisualization.
GraphMining:Laws,Tools,andCaseStudies
D.ChakrabartiandC.Faloutsos
2012
MiningHeterogeneousInformationNetworks:PrinciplesandMethodologies
YizhouSunandJiaweiHan
2012
PrivacyinSocialNetworks
ElenaZheleva,EvimariaTerzi,andLiseGetoor
2012
iii
CommunityDetectionandMininginSocialMedia
LeiTangandHuanLiu
2010
EnsembleMethodsinDataMining:ImprovingAccuracyThroughCombiningPredictions
GiovanniSeniandJohnF.Elder
2010
ModelingandDataMininginBlogosphere
NitinAgarwalandHuanLiu
2009
Copyright© 2012byMorgan&Claypool
Allrightsreserved.Nopartofthispublicationmaybereproduced,storedinaretrievalsystem,ortransmittedin
anyformorbyanymeans—electronic,mechanical,photocopy,recording,oranyotherexceptforbriefquotationsin
printedreviews,withoutthepriorpermissionofthepublisher.
GraphMining:Laws,Tools,andCaseStudies
D.ChakrabartiandC.Faloutsos
www.morganclaypool.com
ISBN:9781608451159 paperback
ISBN:9781608451166 ebook
DOI10.2200/S00449ED1V01Y201209DMK006
APublicationintheMorgan&ClaypoolPublishersseries
SYNTHESISLECTURESONDATAMININGANDKNOWLEDGEDISCOVERY
Lecture#6
SeriesEditors:JiaweiHan,UIUC
LiseGetoor,UniversityofMaryland
WeiWang,UniversityofNorthCarolina,ChapelHill
JohannesGehrke,CornellUniversity
RobertGrossman,UniversityofChicago
SeriesISSN
SynthesisLecturesonDataMiningandKnowledgeDiscovery
Print2151-0067 Electronic2151-0075
Graph Mining
Laws,Tools,andCaseStudies
D.Chakrabarti
Facebook
C.Faloutsos
CMU
SYNTHESISLECTURESONDATAMININGANDKNOWLEDGEDISCOVERY
#6
M
&C Morgan &cLaypool publishers
ABSTRACT
WhatdoestheWeblooklike? Howcanwefindpatterns,communities,outliers,inasocialnetwork?Which
arethemostcentralnodesinanetwork?Thesearethequestionsthatmotivatethiswork.Networksand
graphs appear in many diverse settings,for example in social networks,computer-communication
networks(intrusiondetection,trafficmanagement),protein-proteininteractionnetworksinbiology,
document-textbipartitegraphsintextretrieval,person-accountgraphsinfinancialfrauddetection,
andothers.
Inthiswork,firstwelistseveralsurprisingpatternsthatrealgraphstendtofollow.Thenwe
giveadetailedlistofgeneratorsthattrytomirrorthesepatterns.Generatorsareimportant,because
theycanhelpwith“whatif”scenarios,extrapolations,andanonymization.Thenweprovidealistof
powerfultoolsforgraphanalysis,andspecificallyspectralmethods(SingularValueDecomposition
(SVD)),tensors,andcasestudieslikethefamous“pageRank”algorithmandthe“HITS”algorithm
for ranking web search results.Finally,we conclude with a survey of tools and observations from
relatedfieldslikesociology,whichprovidecomplementaryviewpoints.
KEYWORDS
data mining,social networks,power laws,graph generators,pagerank,singular value
decomposition.
vii
Christos Faloutsos: To Christina, for her patience, support, and down-to-
earth questions; to Michalis and Petros,for the ’99 paper that started it all.
DeepayanChakrabarti:ToPurnaandmyparents,fortheirsupportandhelp,
and for always being there when I needed them.