miss estim mix dattribut data set cost sensit classif data mine cost sensit classif data mine web user brows behavior characterist analysi base bc tree classif algorithm mutual nearest neighbor acquir knowledg inconsist data sourc weight estim semi dparametr miss valu iter imput combin knn imput bootstrap calibr empir likelihood incomplet data analysi summari queri frequent itemset mine cost dsensit classif respect wait cost strategi complex data cube queri confid metric associ rule mine pop algorithm kernel dbase imput treat miss valu knowledg discoveri databas estim confid interv structur differ contrast group miss data mine class dbridg rule base rough set mine global interest pattern multipl databas kernel estim decrement algorithm frequent itemset mainten mine updat databas miss data analysi kernel dbase multi dimput approach secur transact protocol analysi model applic niia nonparametr iter imput algorithm detect differ contrast group mine follow dup correl pattern time drelat databas empir likelihood confid interv differ dataset miss data miss imput base data cluster measur uncertainti differ contrast group identifi depend secur messag protocol analysi cost dsensit imput miss valu order identifi depend secur messag protocol analysi cost dtime sensit decis tree miss valu gbkii imput method miss valu semi dparametr optim miss data imput data structur associ rule discoveri abstract data mine recognis essenti element decis support increasingli focu databas industri computation expens data analysi applic onlin analyt process olap perform kei factor us accept busi critik project client server rule induct technolog industri knowledg acquisit larg databas fund european commiss kind architectur data mine evalu strong focu high perform speci calli data mine techniqu associ rule discoveri decis tree induct implement prototyp present architectur develop critik consortium compar altern architectur edua effici algorithm dynam databas mine editori special issu mine low dqualiti data editori special issu mine low dqualiti data cost dsensit test strategi medic diagnosi doctor determin medic test rai blood test order patient minim total cost medic test misdiagnosi paper design cost sensit machin learn algorithm model learn diagnosi process medic test attribut machin learn valu obtain cost attribut cost misdiagnos misclassif incur cost misclassif cost propos improv decis tree learn algorithm minim sum attribut cost misclassif cost design test strategi request obtain valu unknown attribut cost similar doctor order medic test cost order minim total cost test exampl patient empir evalu compar test strategi show effect outperform previou method case studi heart diseas kernel dbase multi dimput miss data effici frequent itemset mine sampl miss absent question cost dsensit decis tree detect collus attack secur protocol adapt dmean algorithm discov cluster subspac cluster web document base knowledg granular enhanc dwt recent dbias dimens reduct time seri data enhanc dwt recent dbias dimens reduct time seri data gridsm norm ddriven state machin model grid workflow differ detect contrast set identifi follow dcorrel itemset dpair identifi bridg rule conceptu cluster effici comput multi dfeatur data cube jiawei han sonni chee jenni chiang intellig databas system research laboratori school comput scienc simon fraser univers british columbia canada han schee ychiang cs sfu ca url http db cs sfu ca abstract data warehous olap engin expect wide futur data data warehous cleans integr preprocess infrastructur built surround data warehous effici data analysi data warehous olap databas expect major platform data mine futur discuss issu relat effici effect data mine larg data warehous data mart includ desir architectur integr line analyt process olap line analyt mine olam system expect featur olam implement system effect introduct recent develop frequenc decis maker make decis optim paramet miss data imput map relat databas schema owl ontolog deep annot mine multipl data sourc local pattern analysi larg organ process data multipl data sourc branch interst intern compani web emerg larg distribut data repositori consist varieti data sourc format data collect web multipl local dataset bring opportun logic framework identifi qualiti knowledg data sourc norm ddriven state machin model cscw system gener dimens dreduct framework recent dbias time seri analysi abstract recent bias approxim receiv increas attent recent mechan learn trend pattern time seri data stream shown promis cluster time seri increment pattern maintain paper design gener dimens reduct framework recent bias approxim aim make tradit dimensionreduct techniqu action recent bias time seri analysi framework design wai equi segment scheme vari segment scheme scheme time seri data partit segment dimens reduct techniqu appli segment coeffici recent data fewer older data detail preserv recent data fewer coeffici time seri improv effici greatli experiment evalu propos approach demonstr tradit dimens reduct techniqu svd dft dwt pip paa landmark embed framework recent bias approxim stream time seri index term time seri analysi featur extract construct data mine ai advanc artifici intellig ddepth data mine applic stock market ddepth data mine applic stock market osm organiz state machin model cscw system decrement algorithm maintain frequent itemset dynam databas abstract data mine machin learn confront problem pattern mainten data updat fundament oper data manag exist data mine algorithm assum databas static databas updat requir rediscov pattern scan entir data effici mine techniqu data addit databas paper propos decrement algorithm pattern discoveri data delet databas conduct extens experi evalu approach illustr propos algorithm model captur interact data data decreas dynam test dsensit decis tree multipl cost scale recent dbias dimens reduct techniqu time seri data abstract techniqu develop tackl time seri part sequenc equal applic recent data interest signific data paper defin recent bias measur distanc energi propos recent bias techniqu base dwt time seri recent data consid signific recent bias techniqu dimens time seri reduc effect preserv recent bias energi experi demonstr effect propos approach handl time seri identifi interest pattern multidatabas armga identifi interest associ rule genet algorithm partial drepeat approach data mine mine textual document time seri concurr predict movement stock price base content new stori emerg topic data mine commun previou research shown strong relationship time new stori releas time stock price fluctuat paper propos systemat framework predict tertiari movement stock price analyz impact new stori stock specif investig impact new stori stock base effici market hypothesi data mine text mine techniqu extens experi real life data conduct encourag result obtain identifi interest custom web log classif miss miss valu cost dsensit decis tree abstract real world data set machin learn data mine miss valu previou research problem attempt imput miss valu train test paper studi issu costsensit learn consid test cost misclassif cost attribut test expens obtain valu cost effect miss valu similar skip expens riski test miss valu patient diagnosi classif miss miss valu reduc total cost test misclassif meaning imput valu discuss compar strategi util valu miss cost reduct cost sensit decis tree learn index term induct knowledg acquisit machin learn databas classif multi ddatabas mine endl logic framework verifi secur transact protocol verif model electron transact protocol decis tree minim cost propos simpl effect method build test decis tree minim sum misclassif test cost specif put forward origin simpl split criterion attribut select tree build treebuild algorithm desir properti cost sensit learn system account type cost assum test case larg number miss valu design intellig test strategi suggest wai obtain miss valu cost order minim total cost experiment compar strategi demonstr algorithm significantli outperform variat addit algorithm complex similar lower previou work work diagnost task factor misclassif test cost obtain miss inform discov interest associ rule cluster abstract great metric measur interesting rule paper design distinct approach identifi associ rule maxim interesting appli context specif interesting associ rule defin dissimilar cluster addit interesting assist filter rule uninterest applic experi show effect algorithm cost dsensit decis tree multipl cost scale decis tree minim cost propos simpl effect method build test decis tree minim sum misclassif test cost specif put forward origin simpl split criterion attribut select tree build treebuild algorithm desir properti cost sensit learn system account type cost assum test case larg number miss valu design intellig test strategi suggest wai obtain miss valu cost order minim total cost experiment compar strategi demonstr algorithm significantli outperform variat addit algorithm complex similar lower previou work work diagnost task factor misclassif test cost obtain miss inform mine term associ rule heurist queri construct deal inconsist secur messag minimum dsupport identifi larg itemset effici mine posit neg associ rule mine depend pattern probabilist databas guest editor introduct inform enhanc data mine identifi global except pattern multi ddatabas mine multi databas mine local pattern frequent itemset associ rule databas end multi databas mine analyz local pattern gain global pattern put data databas singl dataset destroi import inform reflect distribut global pattern paper develop algorithm synthes local pattern multi databas propos approach fit find potenti except propos method evalu experiment experiment result shown method effici identifi except pattern guest editor introduct inform enhanc data mine fuzzi logic base method acquir user threshold minimum dsupport mine associ rule effici strategi mine except multi ddatabas identifi softwar compon associ genet algorithm verifi purchas request set protocol databas dindepend approach mine associ rule genet algorithm verifi payment author set protocol identifi frequent term text databas associ semant databas dindepend strategi confid determin discov associ larg databas approxim mine dynam databas weight identifi approxim itemset interest larg databas tempor constraint satisfact matrix method constraint satisfact challeng problem interv algebra ia effort attack problem build matrix method deal tempor reason problem develop basic techniqu appli matrix method constraint satisfact paper propag rule algorithm path consist studi matrix method tempor constraint satisfact problem transform problem effect solv data prepar data mine data prepar fundament stage data analysi lot low qualiti inform data sourc web organ compani interest transform data clean form high profit purpos goal gener urgent data analysi aim clean raw data paper show import data prepar data analysi introduc research achiev area data prepar final suggest futur direct research develop agent dbase hybrid framework databas mine knowledg discoveri databas kdd defin iter sequenc step data pre process data mine post data mine signific amount research data mine result varieti algorithm techniqu step singl data mine techniqu proven domain data set techniqu integr hybrid system cooper data mine oper hybrid solut crucial success data mine paper present hybrid framework identifi pattern databas multi databas framework integr techniqu mine task agent point view base experi conduct put kdd techniqu agent base architectur enabl cooper need propos framework highli flexibl robust data mine platform result system demonstr emerg behavior improv perform individu kdd techniqu cooper strategi web data mine clean databas mine pre dprocess collect data multi ddatabas mine multi databas mine import research area urgent analyz data sourc essenti differ mono multi databas mine limit exist multi databas mine effort paper design multidatabas mine process research issu involv mine multi databas includ databas cluster local pattern analysi discuss synthes high dfrequenc rule data sourc post dmine mainten associ rule weight probabl data model semant associ rule mine model algorithm identifi except pattern multi ddatabas mind trend mine increment data mine mine posit neg associ rule paper describ techniqu call apac mine interest quantit associ rule larg databas effect mine rule apac emploi adjust differ analysi techniqu advantag requir user suppli threshold hard determin apac advantag user discov posit neg associ rule posit associ rule tell record attribut attribut neg associ rule tell record attribut attribut fact apac mine posit neg associ rule object meaning measur determin interesting rule make effect data mine task associ rule mine agent mine system emad vision commun data mine agent contribut individu interact decentr control address data mine request emad end user platform research tool chapter detail emad vision conceptu framework current implement emad appli data mine task studi sake breviti concentr agent base associ rule mine agent base classif full descript emad present mine larg databas softwar agent propag tempor relat interv matrix tradit tempor relat propag base allen interv algebra paper propos altern method propag tempor relat interv matric repres tempor relat interv propag tempor relat transform numer comput ef cienci special valu thirteen matric determin tempor relat interv nal result matrix optim propag evalu util propos techniqu implement matrix represent java experiment result demonstr approach ef cient promis real world avoid conquer exploit real world life catch properti time allen propos time world model time interv simpli call interv calculu interv algebra ia allen model divid relationship interv kind tempor form lai foundat deal tempor relationship applic major challeng ia propag tempor relat confront problem tradit model design improv ia number research focus constraint satisfactori discov causal larg databas anytim mine multius applic encod probabl propag belief network collect qualiti data databas mine pattern discoveri probabilist databas logic framework knowledg share multi dagent system abstract issu knowledg share import topic multi agent research knowledg share lead agent analyz judg synthes told inform make agent knowledg match applic paper build logic framework knowledg share agent developa multimod logic reason agent knowledg told inform formal relationshipbetween knowledg told inform present framework semant respect sound complet proof theori mine small databas collect knowledg estim itemset interst sampl larg scale data mine base data partit model compress probabl belief network anytim algorithm agent dmediat merchant inform gather nearest neighborhood algebra probabilist databas tempor logic support histor databas aggreg mainten databas mine imc method interv calculu matrix time represent import applic tempor databas plan multi agent allen work binari interv relaton call interv algebra research investig tempor informaton process base interv calculu limit constraint satisfact np hard problem interv calculu reason propos interpret interv relationship calculu paper establish method transform interv calculu matrix calculu experi show faster propag tempor relat interv algebra keyword plan tempor logic interv algebra interv calculu tempor reason introduct allen propos time world model interv algebra ia model successfulli appli multi agent tempor databas simul concurr sy 