crew cross dmodal resourc search exploit wikipedia integr web resourc wikipedia locat map resourc web abstract map mashup emerg web applic data object blog photo video sourc combin mark map api releas onlin map solut googl yahoo map object typic set tag captur embed semant set coordin indic geograph locat tradit web resourc search strategi effect environ due lack gazett context tag altern approach locat object tag match number tag object typic small make difficult object captur complet semant queri object paper focu fundament applic locat geograph resourc propos effici tagcentr queri process strategi aim find set nearest locat object match queri tag fact larg number data object tag develop effici search algorithm scale term number object tag ensur result relev propos geograph context sensit geo tf idf rank mechan experi synthet data set demonstr scalabl experi real life data set confirm practic call order hierarch brows approach elicit user prefer multipl featur fusion social media applic effici effect similar search probabilist data base earth mover distanc triangul dbase dens neighborhood graph discoveri understand mean shift sky gener framework extend skylin queri skylin queri data set multi dimension space decis make applic tradition object domin object dimens wors dimens skylin data set consist object domin object cater applic requir control size skylin handl data set structur work propos extend definit skylin base variant domin relationship view prolifer variant paper gener framework propos guid extens skylin queri convent definit variant framework explicitli carefulli examin properti preserv vari find time dlag cluster abstract exist cluster algorithm gene sampl time express data time lag correl gene express pattern ignor correl time subseri disregard continu time seri valid pure shift pure scale coher pattern gener shiftingand scale pattern paper propos cluster model cluster address problem reflect shift scale correl dimension gene sampl time data cluster model express level gene shift scale coher sampl subspac time subseri arbitrari time lag develop cluster algorithm lagmin identifi interest cluster satisfi constraint regul coher minimum gene number ming minimum sampl subspac size min minimum time period length mint experiment result synthet real life dataset show lagmin effect scalabl paramet robust gene express data paper model algorithm appli data spatial tempor coher pursu keyword search spatial databas search document abstract work address spatial keyword queri call closest keyword mck queri databas spatial object tupl descript minim commun cost continu skylin mainten exist work skylin literatur focus optim process cost paper aim minim commun overhead client server architectur server continu maintain skylin dynam object contribut filter method avoid transmiss updat object influenc skylin specif object assign filter issu updat violat filter filter achiev signific save naiv approach transmit updat step introduc concept frequent skylin queri slide window fsqw motiv snapshot skylin stream environ chang time fsqw report object skylin recent timestamp lt filter easili adapt fsqw process potenti high overhead larg frequent updat dataset reduc commun cost propos sampl method return approxim fsqw result comput snapshot skylin final integr filter sampl hybrid approach combin individu advantag minim commun cost continu skylin mainten exist work skylin literatur focus optim process cost paper aim minim commun overhead client server architectur server continu maintain skylin dynam object contribut filter method avoid transmiss updat object influenc skylin specif object assign filter issu updat violat filter filter achiev signific save naiv approach transmit updat step introduc concept frequent skylin queri slide window fsqw motiv snapshot skylin stream environ chang time fsqw report object skylin recent timestamp lt filter easili adapt fsqw process potenti high overhead larg frequent updat dataset reduc commun cost propos sampl method return approxim fsqw result comput snapshot skylin final integr filter sampl hybrid approach combin individu advantag kernel dbase skylin cardin estim skylin dimension dataset consist point domin incorpor skylin oper practic databas system necessit effici effect cardin estim modul exist theoret work problem limit case dimens independ rare hold real dataset state art log sampl ls techniqu simpli appli theoret result independ dimens independ data lead larg estim error solv problem propos kernel base kb approach approxim skylin cardin nonparametr method extens experi real dataset demonstr kb achiev high accuraci case ls fail time numer natur effici kb compar ls extend ls kb domin skylin commonli convent skylin high dimension data rule dbase classif hniqu sell point fuzzi system interpret rule base uccutuc rule base measur compar fuzzi system solut suggest index measurz interpret fuzzi rule base classif problem index describ interpret rule base system set partit variabl demonstr featur index data set simpl benchmark set real world similar search bregman diverg dmetric index paper examin problem index metric distanc function focu gener class distanc function bregman diverg support nearest neighbor rang queri distanc function kl diverg itakura saito distanc special case bregman diverg wide applic statist speech recognit time seri analysi unlik metric space kei properti triangl inequ distanc symmetri hold distanc function direct adapt exist index infrastructur develop metric space devis solut handl class distanc measur expand map point origin space extend space subsequ show state art tree base index method low moder dimension dataset vector approxim file va file method high dimension dataset adapt extend space answer queri effici improv distanc bound techniqu distribut base index optim introduc improv perform queri answer index construct appli tree va file extens experi conduct valid approach varieti dataset rang bregman diverg function compar star approxim graph edit distanc graph data ubiquit manipul base similar essenti applic graph edit distanc wide accept measur determin similar graph extens applic field pattern recognit comput vision problem graph edit distanc comput np hard gener paper introduc method comput upper lower bound edit distanc graph polynomi time appli method algorithm appful appsub introduc perform kind graph search graph databas comprehens experiment studi conduct real synthet dataset examin aspect method bound graph edit distanc result show method achiev good scalabl term number graph size graph effect algorithm confirm us bound filter search graph sens physic abstract paper describ implement framework integr knowledg represent reason symbiot system system number heterogen sensor pervas embed environ mobil robot human exist commun work integr mediat perceptu anchor creat maintain correspond symbol system perceptu data refer physic object framework evalu researchcyc knowledg represent reason system context physic testb consist small apart home domin game analysi microeconom data mine game theori power tool analysi competit manufactur market paper present studi combin game theori data mine introduc concept domin game analysi present multidimension market model dimens repres attribut commod product custom repres point multidimension space product domin custom attribut satisfi requir custom expect market share product measur expect number buyer custom equal bui product domin nash equilibrium configur product achiev stabl expect market share product prove nash equilibrium model comput polynomi time manufactur modifi product round robin manner improv effici comput design algorithm manufactur effici find respons product market skyfram framework skylin queri process peer dto dpeer system uncertain cluster model method applic revisit cluster problem certainti uncertainti discov spatial interact pattern tutori articl submit public journal consider commiss organ report repres idea author offici view school univers discuss content report author address shown cover valid multi dcolumn schema match type abstract valid multi column schema match essenti success databas integr task difficult databas integr overlap data case practic custom base compani base intuit valu present column relat schema match similar semant type captur distribut valu statist type develop method valid composit schema match techniqu base kei technic idea propos gener measur compar column match schema match base notion inform theoret discrep gener standard geometr discrep basi match present algorithm split string valu column identifi substr match valu column enabl multi column schema match techniqu invalid certif fail valid schema match complement conceptu algorithm contribut experiment studi demonstr effect effici techniqu varieti databas schema data set queri complex spatio dtempor sequenc human motion databas abstract content base retriev spatio tempor pattern human motion databas inher nontrivi find effect distanc measur data difficult data typic model time seri high dimension vector incur expens storag retriev cost result high dimension paper abstract complex spatio tempor data set frame repres high dimension categor featur vector distanc measur queri high dimension categor time seri propos effici queri process techniqu answer queri develop conduct experi propos distanc measur queri human motion captur databas result signific improv effici queri process categor time seri time faster origin motion sequenc achiev guarante effect search skylin djoin distribut databas abstract checkpoint rollback recoveri techniqu handl failur distribut databas system paper establish suffici condit checkpoint set data item part transact consist global checkpoint distribut databas throw light design effici intrus checkpoint techniqu transpar recoveri techniqu distribut databas system skylin djoin distribut databas abstract checkpoint rollback recoveri techniqu handl failur distribut databas system paper establish suffici condit checkpoint set data item part transact consist global checkpoint distribut databas throw light design effici intrus checkpoint techniqu transpar recoveri techniqu distribut databas system estim local optimum em algorithm gaussian mixtur model em algorithm popular iterationbas method estim paramet gaussian mixtur model larg observ set case em algorithm guarante converg global optimum stop local optimum wors global optimum requir run multipl procedur em algorithm initi configur return solut improv effici scheme propos method estim upper bound logarithm likelihood local optimum base current configur latest em iter accomplish deriv region bound locat local optimum upper bound estim maximum likelihood estim termin em algorithm procedur estim local optimum wors solut extens experi show method effect effici acceler convent multipl restart em algorithm categor skylin stream data problem skylin comput attract consider research attent categor domain problem complic primarili due partial order natur attribut tupl paper initi studi stream categor skylin identifi limit exist work offlin categor skylin comput realiz techniqu problem maintain skylin categor data stream environ develop lightweight data structur index tupl stream buffer gracefulli adapt tupl attribut partial order domain size complex addition studi domin relat dual space util geometr arrang order index categor skylin effici evalu domin queri lastli experiment studi evalu effici propos techniqu csv visual mine cohes subgraph extract dens compon graph effici import object wide rang applic domain rang social network analysi biolog network analysi world wide web stock market analysi motiv recent algorithm tackl problem base frequent pattern mine paradigm limit method highli sensit paramet set reli exhaust enumer exponenti time complex fail user understand underli distribut compon embed host graph articl propos approxim algorithm mine visual cohes subgraph dens compon larg graph approach refere cohes subgraph visual csv reli map strategi map edg node multidimension space dens area map space correspond cohes subgraph algorithm walk dens region map space output visual plot effect captur dens subcompon distribut graph unlik extant algorithm exponenti complex csv complex logv fix paramet map dimens correspond number vertic graph real dataset perform typic quadrat demonstr util csv stand tool visual graph explor pre filter step significantli scale exact subgraph mine algorithm graph method keyword dbase select top dk databas databas manag system offer comprehens solut data storag requir deep knowledg schema data manipul languag order perform effect retriev requir pose problem lai occasion user method incorpor keyword search ks relat databas exist techniqu focu queri singl dbm hand prolifer distribut databas convent emerg applic necessit support keyword base data share queri multipl dmbss order avoid high cost search numer potenti irrelev databas system propos ks method select top candid base potenti result queri ks summar databas keyword relationship graph node repres term edg describ relationship keyword relationship graph util comput similar databas ks queri queri process promis databas search extens experiment evalu demonstr ks outperform current state art techniqu aspect includ precis recal effici space overhead flexibl accommod semant continu dmean monitor move object approach dynam dtune databas buffer current busi reli heavili effici access databas manual tune databas system perform expert increasingli infeas small compani hire expert expens larg enterpris expert fulli understand interact larg system multipl chang workload trend led major vendor offer tool automat dynam tune databas system databas tune knob concern buffer pool cach data disk page specif knob control buffer alloc cach miss probabl direct impact perform previou method automat buffer tune base simul black box control gradient descent empir equat articl present approach calcul analyt deriv equat relat miss probabl buffer alloc equat fit buffer replac polici twelv dataset mainfram run commerci databas larg corpor equat identifi buffer size limit buffer tune power idl buffer replac simul predict cost experi postgresql effici skylin queri process peer dto dpeer network skylin queri gain interest databas research commun recent year exist studi focu central system resolv problem distribut environ peer peer network emerg topic desiderata effici skylin queri environ includ progress return answer low process cost term number peer access search messag balanc queri load peer paper propos solut satisfi desiderata solut base balanc tree structur network partit skylin search space adapt base queri access pattern allevi problem hot spot present skylin queri process estim peer node queri subspac control amount queri forward limit number peer involv amount messag transmit network load balanc achiev queri load consciou data space split merg join departur node dynam load migrat experi real synthet dataset confirm effect scalabl algorithm network spade shape dbase pattern detect stream time seri monitor predefin pattern stream time seri applic trend relat analysi sensor network video surveil current studi monitor emploi euclidean distanc calcul similar queri pattern subsequ stream time seri euclidean distanc shown ineffect measur distanc time seri shift scale exist warp distanc dynam time warp dtw longest common subsequ lcss propos handl warp tempor dimens inadequ handl shift scale amplitud dimens design full sequenc match onlin monitor applic typic knowledg posit length match subsequ paper discuss weak exist warp distanc detect pattern stream time seri propos warp distanc spatial assembl distanc spade handl shift scale tempor amplitud dimens propos effici approach continu pattern detect spade fundament subsequ match stream data final experiment result show spade effect effici continu pattern detect stream time seri effect keyword dbase select relat databas world wide web fuel demand incorpor keyword base search structur databas current research work focus keywordbas search singl structur data sourc grow interest distribut databas servic orient architectur internet import extend capabl multipl structur data sourc import problem enabl queri facil select data sourc relev keyword queri tradit databas summari techniqu select unstructur data sourc develop ir literatur inadequ problem captur structur data sourc paper studi databas select problem relat data sourc propos method effect summar relationship keyword relat databas base structur develop effect rank method base keyword relationship summari order select databas keyword queri implement system planetlab environ extens experi real dataset demonstr effect propos summar method effici process subspac skylin queri high dimension data recent studi effici answer subspac skylin queri separ approach focus pre materi set skylin point subspac focu dynam answer queri set anchor prune skylin point spatial reason effort compress pre materi subspac skylin remov redund storag space approach remain exponenti number dimens queri time approach hand grow substanti data higher dimension prune power anchor weaker paper propos method answer subspac skylin queri high dimension data premateri storag queri time moder propos notion maxim partial domin space maxim partial domin space maxim equal space pair skylin object full space concept foundat answer subspac skylin queri high dimension data queri process involv simpl prune oper skylin comput small subset candid skylin point subspac develop random sampl method comput subspac skylin line fashion extens experi conduct demonstr effici effect method domin neighborhood profit recent research skylin queri attract interest databas data mine commun databas object belong skylin domin respect attribut databas object current method consid call min max attribut price qualiti user minim maxim object spatial attribut coordin repres relev constraint queri result paper introduc skylin queri type take account min max attribut spatial attribut relationship attribut type queri support micro econom approach decis make qualiti cost solut investig altern approach effici queri process symmetr base shelf index structur asymmetr base index structur special purpos extens experiment evalu real dataset synthet dataset demonstr queri type meaning propos algorithm effici scalabl compress hierarch mine frequent close pattern dens data set proceed intern confer manag data mine outlier spatial network abstract outlier analysi import task data mine attract attent research applic previou work outlier detect involv type databas spatial databas time seri databas biomed databas exist studi consid spatial network point resid edg paper studi interest problem distanc base outlier spatial network propos effici mine method partit edg spatial network set length segment quickli identifi outlier remain edg prune unnecessari edg outlier present algorithm appli spatial network updat point input paramet outlier measur chang experiment result verifi scalabl effici propos method high dimension skylin decis make applic skylin queri frequent find set domin data point call skylin point multidimension dataset high dimension space skylin point longer offer interest insight paper introduc metric call skylin frequenc compar rank interesting data point base return skylin number dimens subspac consid intuit point high skylin frequenc interest domin fewer combin dimens problem find top frequent skylin point algorithm propos skylin comput typic scale dimension frequent skylin comput requir skylin comput exponenti number subset dimens present effici approxim algorithm address twin difficulti extens perform studi show approxim algorithm run fast comput correct result larg data set high dimension space mine shift dand dscale dregul pattern gene express profil paper propos model coher cluster gene express data call reg cluster propos model express profil gene cluster follow shift scale pattern subspac scale posit neg express condit cluster signific previou work measur task set densiti base subspac cluster algorithm requir gene similar express level subspac pattern base biclust algorithmsonlyallowpureshiftingorpurescalingpattern tendenc base biclust algorithm coher guarante develop patternbas biclust algorithm identifi shift andscal regul pattern satisfi coher constraint regul constraint experiment result show reg cluster algorithm detect signific amount cluster miss previou model cluster potenti high biolog signific lower bound local optimum dmean algorithm mean algorithm popular cluster method field comput scienc data mine machin learn inform retriev mean algorithm converg local optimum wors desir global optim solut overcom problem current mean algorithm variant run time initi center avoid trap local optimum unaccept qualiti paper propos effici method comput lower bound cost local optimum current center set mean iter mean algorithm halt procedur lower bound cost futur local optimum wors solut comput lower bound comput incur extra time consumpt iter extens experi synthet real data set show method greatli prune unnecessari iter improv effici algorithm data set high dimension larg opportun xml data mine modern applic rank outlier symmetr neighborhood relationship abstract mine outlier databas find except object deviat rest data set classic outlier analysi algorithm recent studi focus mine local outlier outlier densiti distribut significantli neighborhood estim densiti distribut locat object base densiti distribut nearest neighbor outlier locat densiti distribut neighborhood significantli case object spars cluster close denser cluster result wrong estim avoid problem propos simpl effect measur local outlier base symmetr neighborhood relationship propos measur consid neighbor revers neighbor object estim densiti distribut result outlier discov meaning comput local outlier effici mine algorithm develop detect top outlier base definit comprehens perform evalu analysi show method effici comput effect rank outlier dada data cube domin relationship analysi concept domin recent attract interest context skylin comput ndimension data set apointp domin dimens equal remain dimens paper propos extend concept domin busi analysi microeconom perspect specif propos form analysi call domin relationship analysi dra aim provid insight domin relationship product potenti buyer analyz relationship compani posit product effect remain profit support dra propos data cube call dada data cube domin relationship analysi captur domin relationship product custom type queri call domin relationship queri drq propos analysi purpos linear optim queri loq subspac analysi queri saq compar domin queri cdq algorithm design effici comput dada answer drq dada result comprehens experi show effect effici dada queri process strategi find ddomin skylin high dimension space dada data cube domin relationship analysi concept domin recent attract interest context skylin comput ndimension data set apointp domin dimens equal remain dimens paper propos extend concept domin busi analysi microeconom perspect specif propos form analysi call domin relationship analysi dra aim provid insight domin relationship product potenti buyer analyz relationship compani posit product effect remain profit support dra propos data cube call dada data cube domin relationship analysi captur domin relationship product custom type queri call domin relationship queri drq propos analysi purpos linear optim queri loq subspac analysi queri saq compar domin queri cdq algorithm design effici comput dada answer drq dada result comprehens experi show effect effici dada queri process strategi mine frequent sequenti pattern similar search match base approach similar search crucial task multimedia retriev data mine exist work model problem nearest neighbor nn problem consid distanc queri object data object fix set featur approach drawback leav partial similar uncov distanc affect dimens high dissimilar overcom drawback propos match problem paper match problem model similar search match queri object data object dimens integ smaller dimension dimens determin dynam make queri object data object return answer set match match queri expect mine frequent close cube dataset paper introduc concept frequent close cube fcc gener notion frequent close pattern context propos algorithm mine fcc dataset scheme repres slice mine rsm framework extend exist fcp mine algorithm fcc mine techniqu call cubemin algorithm oper space directli implement scheme evalu perform real synthet dataset experiment result show rsm base scheme effici dimens small cubemin superior scalabl cluster graphic processor arc aggreg relat column score scheme align sequenc motiv biologist frequent align multipl biolog sequenc determin consensu sequenc search predomin residu conserv region determin conserv region align import activ protein sequenc hundr residu longer difficult distinguish biolog import conserv region motif domain wide tool logo al confind entropi base method fail highlight region comput tool highlight biolog import region accur highli desir result paper present score scheme arc aggreg relat column score align biolog sequenc arc method consid tradit charact similar measur column correl extens experiment evalu prosit pattern arc highlight motif region accuraci top peak avail sourc code continu skylin queri move object impact doubl blind review sigmod public detail analysi discov strong skylin point high dimension space dsim distanc dbase index method genom sequenc discov strong skylin point high dimension space index dna sequenc dgram abstract observ recent year grow interest similar search larg collect biolog sequenc contribut interest paper present method index dna sequenc effici base gram facilit similar search dna databas sidestep linear scan entir databas level index hash tabl tree propos base gram dna sequenc propos data structur quick detect sequenc distanc queri sequenc experiment result show method effici detect similar region dna sequenc databas high sensit data mine techniqu microarrai dataset find repres set massiv data inform ag data pervas applic data explos signific phenomenon massiv data volum pose challeng human user comput project propos model identifi repres set larg databas repres set special subset origin dataset main characterist significantli smaller size compar origin dataset captur inform origin dataset compar subset size low redund repres informationtheoret measur mutual inform rel entropi measur repres repres set design greedi algorithm present heurist algorithm deliv perform run experi real dataset evalu effect repres set term coverag accuraci experi show repres set attain expect characterist captur inform effici knr dtree dtree dbase index facilit spatial window queri relat spatial relat mobil environ tree base index facilit curler find visual nonlinear correl cluster work find linear correl subset featur high dimension data work detect nonlinear correl left larg untouch paper present algorithm find visual nonlinear correl cluster subspac high dimension databas unlik detect linear correl cluster uniqu orient find nonlinear correl cluster vari orient requir merg cluster possibl orient combin fact spatial proxim judg base subset featur origin decid cluster merg cluster process challeng avoid problem propos concept call share level captur spatial proxim cluster orient judg similar cluster base concept develop algorithm detect nonlinear correl cluster visual experi synthet real life dataset show effect method mine top dk cover rule group gene express data similar evalu tree dstructur data index mix type approxim retriev applic data cleans retriev categor numer attribut base notion approxim match edit distanc numer distanc profound import commonli approxim match predic combin attribut conjunct exist databas techniqu approxim retriev limit applic singl attribut retriev tree variant paper propos methodolog util multidimension index structur problem approxim multi attribut retriev method enabl index collect string numer attribut facilit approxim retriev edit distanc approxim match predic string numer distanc numer attribut approach present base repres set string higher level index structur suitabl compress reason edit distanc queri string compress trie index node feasibl propos evalu techniqu gener compress trie represent fulli index methodolog experiment result show benefit propos compar altern strategi problem fast filter obstruct nearest neighbor queri local signatur tabl fast similar search transact data green recycl reus frequent pattern ldc enabl search partial distanc hyper ddimension space recent advanc research field multimedia bioinformat brought gener hyper dimension databas hundr thousand dimens hyperdimension databas pose signific problem exist high dimension index techniqu develop index databas commonli hundr dimens support cient queri retriev hyper dimension databas propos methodolog call local digit code ldc support nearest neighbor knn queri hyper dimension databas exist ubiquit indic tree ldc extract simpl bitmap represent call digit code dc point databas prune knn search perform dynam select subset bit dc base subsequ comparison perform expens oper involv comput norm distanc function hyper dimension data avoid extens experi conduct show methodolog er signific perform advantag exist index method real life synthet hyper dimension dataset substructur cluster sequenti object dataset sequenti object sequenti object set point locat dimension space link form sequenc set sequenti object aim find significantli larg substructur present sequenti object unlik tradit subspac cluster method object compar base valu dimens match dimens sequenti object affect translat rotat object defin similar object judg comput structur distanc measur call requir proper align includ translat rotat object comput expens propos measur call amp shown experiemnt approxim base defin cluster model call devis algorithm discov maximum sequenti dataset experi conduct illustr effici effect algorithm itcompress iter semant compress algorithm real dataset larg necessit data compress tradit syntact data compress method treat tabl larg byte string oper byte level tradeoff case eas retriev eas retriev singl tupl attribut decompress larger unit effect compress regard semant compress gener consider interest motiv recent work paper propos semant compress algorithm call itcompress iter compress achiev good compress permit access attribut level requir decompress larger unit itcompress iter improv compress ratio compress output scan tabl amount compress tune base number iter initi iter provid signific compress make cost effect compress techniqu extens experi conduct result superior itcompress respect previous tehniqu spartan fascicl mine frequent close pattern microarrai data microarrai data typic larg number column small number row pose great challeng exist frequent close pattern mine algorithm discov pattern item enumer space paper propos algorithm explor row enumer space mine frequent close pattern experi real life gene express data show algorithm faster exist algorithm includ closet charm closet car penter increment mainten quotient cube median data cube pre comput import concept support olap onlin analyt process studi extens feasibl comput complet data cube due huge storag requir recent propos quotient cube address issu partit method group cube cell equival partit approach distribut aggreg function sum appli holist aggreg function median maintain data cube holist aggreg hard problem difficulti li fact histori tupl valu order comput aggreg tupl insert delet quotient cube make problem harder maintain equival class paper introduc techniqu call addset data structur slide window deal problem develop effici algorithm maintain quotient cube holist aggreg function take small storag space perform studi show algorithm effect effici scalabl larg databas farmer find interest rule group microarrai dataset microarrai dataset typic larg number column small number row associ rule prove analyz dataset exist associ rule mine algorithm unabl cientli handl dataset larg number column number associ rule gener dataset enorm due larg number column combin cobbler combin column row enumer close pattern discoveri problem mine frequent close pattern receiv consider attent recent promis redund compar discov frequent pattern exist algorithm present separ group featur column enumer row enumer featur enumer algorithm charm closet effici dataset small number featur larg number row number featur combin enumer small row enumer algorithm carpent hand suitabl dataset bioinformat data larg number featur small number row group algorithm encount problem dataset larg number row featur paper describ algorithm call cob bler effici mine dataset cobbler design dynam switch featur enumer row enumer depend data characterist process mine portion dataset process suitabl method make mine effici experi real life synthet dataset show cobbler order magnitud previou close pattern mine algorithm charm closet carpent string join preced count matrix semant mine analysi gene express data associ rule reveal biolog relev relationship gene environ categori exist associ rule mine algorithm render impract gene express data typic thousand ten thousand column gene express level ten row sampl main problem algorithm exponenti depend number column shortcom evid associ gener kind data end develop depth row wise algorithm farmer special design cientli discov cluster associ rule interest rule group irg satisfi user minimum support confid chi squar threshold biolog dataset oppos find associ rule individu base farmer develop prototyp system integr semant mine visual analysi irg mine gene express data pier effici model similar search dna sequenc databas grow interest genom research result creation huge biolog sequenc databas paper present hash base pier model effici homolog search larg dna sequenc databas model segment databas call pier access search opposit approach requir full scan biolog sequenc databas improv search effici pier store special design hash tabl help avoid expens align oper hash tabl small resid main memori avoid search step show theoret empir propos approach effici detect biolog sequenc similar queri sequenc high sensit search mine retriev histori content dbase imag retriev carpent find close pattern long biolog dataset characterist dataset typic larg number column small number row gene express dataset column row ed dtree index larg dna sequenc databas grow interest genom research caus explos growth size dna databas make increas challeng perform search paper propos index structur call ed tree support fast effect homolog search dna databas ed tree develop enabl probe base homolog search algorithm blastn gener short probe string queri sequenc match sequenc databas order identifi potenti region high similar queri sequenc unlik blastn homolog search algorithm develop ed tree support flexibl probe model longer probe relax match consequ ed tree effect effici latest blastn ncbi blast support homolog search take moder storag compar exist data structur suffix tree index dna databas giga base pair gbp ed tree take gb secondari storag easili handl desktop pc experi shown paper support claim effici mine intertransact associ rule fault dtoler frequent pattern mine problem challeng spatial cluster presenc obstacl cluster spatial data mine group similar object base distanc connect rel densiti space real world exist physic obstacl river lake high wai presenc affect result clu tere substanti paper studi problem cluster presenc obstacl defin cod cluster obstruct distanc prob lem solut problem propos scal cluster algorithm call cod claran discuss form pre process inform enhanc efici cod claran strictest sens codproblem treat chang distanc function handl current cluster algorithm chang di tanc function show push task handl obstacl cod claran abstract distanc function level op timiz form prune func tion conduct perform studi show cod claran efici effec tive constraint dbase cluster larg databas constrain cluster nding cluster satisfi user speci ed constraint highli desir applic paper introduc constrain cluster problem show tradit cluster algorithm mean handl scalabl constraint cluster algorithm coca develop studi start nding initi solut sati es user speci ed constraint ne solut perform con ned object movement constraint algorithm consist phase pivot movement deadlock resolut phase show nding optim solut np hard propos heurist show algorithm scale larg data set heurist micro cluster share experi show eectiv ecienc heurist mine top dn local outlier larg databas outlier detect import task data mine numer applic includ credit card fraud detect video surveil recent work outlier detect introduc notion local outlier degre object outli depend densiti local neighborhood object assign local outlier factor lof repres likelihood object outlier concept local outlier comput lofvalu data object requir larg number nearest neighbor search computation expens object outlier provid user option find outstand local outlier top data object local outlier lof prune carefulli find top outlier result amount comput find loffor object paper propos method effici find top local outlier larg databas concept micro cluster introduc compress data effici micro cluster base local outlier mine algorithm design base concept algorithm advers affect overlap micro cluster propos meaning cut plane solut overlap data formal analysi experi show method achiev good perform find outstand local outlier geo dspatial cluster user dspecifi constraint duminda wijesekera daniel barbara session discuss variat multimedia data mine simeon simoff discuss multimedia mine encompass open issu coffe break sigkdd confer open award viii ix tabl content geo spatial cluster user constraint anthoni tong raymond ng lak lakshmanan jiawei han multi level index gi enhanc learn satellit imageri krzysztof koperski giovanni marchisio predict locat map similar plum framework spatial data mine sanjai chawla shashi shekhar weili wu uygar ozesmi learn prosod pattern mandarin speech synthesi yiqiang chen wen gao tingshao zhu coe cluster obstacl entiti preliminari studi moo methodolog onlin optim mine offlin optimum break barrier transact mine inter dtransact associ rule broom buffer replac onlin optim mine report interest techniqu discov associ rule problem mine associ rule introduc supermarket supermarket bar code technolog made collect call basket data store item purchas transact basi associ rule mine basket data transact bread butter transact purchas milk buffer manag distribut databas system data mine base approach paper propos data mine base approach public buffer manag distribut databas system databas buffer organ area public privat privat buffer area page updat user public buffer area page share user site tradit buffer manag strategi limit knowledg user access pattern propos approach discov knowledg page access sequenc user transact guid public buffer placement replac knowledg discov discoveri algorithm discuss effect propos approach investig simul studi result discov knowledg public buffer hit ratio improv significantli introduct due higher cost fetch data disk ram databas 