empir studi categori skew featur select text categor abstract paper present empir comparison effect categori skew featur select method method evalu dataset gener newsgroup ohsum reuter text corpora dataset gener possess categori skew characterist number document assign categori object determin perform featur select method measur measur precis number featur need produc perform found highest measur valu obtain bi normal separ inform gain highest precis valu obtain categor proport differ chi squar categor proport differ featur select method text categor supervis text categor machin learn task predefin categori label automat assign previous unlabel document base characterist word contain document number uniqu word learn task number featur larg effici accuraci learn task increas featur select method extract document subset featur consid relev paper introduc featur select method call categor proport differ cpd measur degre word contribut differenti categori categori cpd word categori text corpu ratio consid number document categori word occur number document categori word occur conduct seri experi evalu cpd conjunct svm naiv bay text classifi ohsum newsgroup reuter text corpora recal precis measur measur perform result obtain cpd compar obtain common featur select method found literatur inform gain document frequenc mutual inform odd ratio simplifi empir result show gener measur cpd outperform featur select method text categor task pattern classif dlimit poker head dstart evolutionari approach evalu domin dbase rough set approach interfac design exploratori quantit contrast set mine discret approach mine associ interfac design abstract consum research consum compensatori compensatori decis strategi formul purchas decis compensatori decis make strategi consum fulli ration decis outcom compensatori decis make strategi consum consid inform mean time decis design onlin shop support tool incorpor decis make strategi goal person design user interfac enhanc qualiti satisfact consum shop experi paper present work goal author describ research refin previous develop procedur techniqu cluster analysi rough set obtain consum inform need support design customiz person user interfac enhanc author refin procedur examin evalu techniqu tradit associ mine specif conduct experiment eclat algorithm author previou work summari discuss previou work relat evalu provid result analyz opportun futur work statist methodolog mine potenti interest contrast set evalu wordnet featur text classif model character wordnet featur boolean model text classif evalu wordnet featur text classif model predict itemset sale profil share measur repeat dbui theori random sampl sale transact record scanner panel period week month quarter analyz scanner panel determin approxim penetr purchas frequenc distribut frequent purchas item itemset purchas frequenc distribut item itemset current period model neg binomi distribut paramet model predict sale profil period present repres experiment result base synthet data measur interesting discov knowledg principl approach evalu interesting measur rank discov knowledg mine larg databas number pattern discov easili exce capabl human user identifi interest result address problem techniqu suggest reduc order pattern prior present user paper focu rank summari gener singl dataset attribut gener wai level granular taxonom hierarchi theoret empir evalu thirteen divers measur heurist measur interesting rank summari gener databas thirteen divers measur previous util disciplin inform theori statist ecolog econom describ principl measur satisfi consid rank summari theoret result show thirteen divers measur satisfi principl analyz distribut index valu gener thirteen divers measur empir result obtain synthet data show distribut index valu gener tend highli skew median middl index valu object work gain insight behaviour expect measur practic lorenz domin order measur interesting kdd abstract rank summari gener databas context descript data mine task singl data set gener wai level granular approach gener summari base data structur attribut call domain gener graph dgg dgg attribut direct graph node repres domain valu creat partit origin domain attribut edg repres gener relat domain set dgg set attribut gener space defin combin domain domain select dgg combin gener space describ summari consist dgg gener select attribut number attribut gener larg dgg attribut complex gener space larg result gener summari number summari easili exce capabl domain expert identifi interest result paper show lorenz domin order rank summari prior present domain expert lorenz domin order defin partial order summari case case defin total order rank order summari repres object evalu rel interesting domain expert start point subject evalu summari evalu interesting measur rank discov knowledg mine larg databas number pattern discov easili exce capabl human user identifi interest result address problem techniqu suggest reduc order pattern prior present user paper focu rank summari gener singl dataset attribut gener wai level granular taxonom hierarchi theoret empir evalu thirteen divers measur heurist measur interesting rank summari gener databas thirteen divers measur previous util disciplin inform theori statist ecolog econom describ principl measur satisfi consid rank summari theoret result show thirteen divers measur satisfi principl analyz distribut index valu gener thirteen divers measur empir result obtain synthet data show distribut index valu gener tend highli skew median middl index valu object work gain insight behaviour expect measur practic principl mine summari object measur interesting import problem area data mine develop effect measur interesting rank discov knowledg paper propos principl measur satisfi consid rank interesting summari gener databas investig problem context summar singl dataset gener wai level granular perform compar sensit analysi fifteen divers measur identifi satisfi propos principl fifteen divers measur previous util disciplin inform theori statist ecolog econom object measur interesting rank summari gener databas object work gain insight behaviour expect divers measur practic begin develop theori interesting util measur assess appli object interesting measur data mine system import step knowledg discoveri task interpret evalu discov pattern address problem techniqu chi squar test independ suggest reduc number pattern present user focu attent statist signific mine larg databas number pattern discov remain larg adjust signific threshold elimin spuriou pattern need effect measur assist interpret evalu step rank interesting remain pattern prior present user paper describ step process rank interesting discov pattern util chi squar test independ step object measur interesting step show step process appli rank character gener associ rule data cube rank interesting summari data mine system studi data mine task descript summar represent languag gener relat evalu criteria base heurist measur interesting method search multi attribut gener algorithm domain gener graph present empir compar heurist rank interesting gener relat summari measur base common measur divers popul statist varianc simpson index shannon index measur rank complex summari tupl attribut interest highli rank summari provid reason start point analysi discov knowledg heurist rank interestig discov knowledg heurist measur interesting tupl gener relat summari gener databas uniqu consid popul structur probabl distribut paper present empir compar sixteen heurist measur evalu structur summari assign singl real valu index repres interesting rel summari gener databas heurist base measur divers dispers domin inequ area physic social ecolog manag inform comput scienc rank summari gener databas applic area sixteen heurist rank complex summari tupl attribut interest demonstr sampl data set order measur rank summari highli correl tempor gener domain gener graph paper address problem domain gener graph gener tempor data extract relat databas domain gener graph attribut defin partial order repres set gener relat attribut propos formal specif domain gener graph calendar date time attribut graph reusabl gener calendar attribut adapt extend restrict applic transport databas calendar attribut data mine larg databas domain gener graph attribut orient gener summar inform relat databas repeatedli replac specif attribut valu gener concept userdefin concept hierarchi introduc domain gener graph control gener set attribut show construct present serial parallel version multi attribut gener algorithm travers gener state space join domain gener graph multipl attribut base gener test approach algorithm gener summari consist domain gener graph experiment result show signific speedup partit path combin dgg multipl processor rank interesting result summari measur base varianc rel entropi experiment result show measur provid effect basi analyz summari data gener relat databas varianc appear rank complex summari attribut tupl interest techniqu gener tempor durat relat databas address problem gener tempor data durat extract relat databas approach base domain gener graph defin partial order gener relat durat attribut domain gener graph reusabl gener durat attribut adapt extend restrict applic transport databas durat attribut mine market basket data share measur character itemset propos share confid framework knowledg discoveri databas address problem mine itemset market basket data goal fold present itemset measur practic altern commonli support measur discov bui pattern custom discov custom profil partit custom distinct class present algorithm classifi itemset base characterist attribut extract censu lifestyl data algorithm combin apriori algorithm discov associ rule item larg databas aog algorithm attribut orient gener larg databas suggest character itemset gener concept hierarchi characterist attribut final present experiment result demonstr util shareconfid framework gener lattic gener lattic encod domain knowledg relev gener provid conveni framework data visual user guid explor autom guidanc independ explor reduc size gener lattic individu attribut defin type prune gener space defin cross product lattic attribut increas relev data explor result defin addit type prune interact webbas system visual gener space user interact guid data explor process gener calendar attribut domain gener graph paper address problem gener tempor data base calendar date time attribut propos method base domain gener graph lattic defin partial order repres set gener relat attribut compon domain gener graph suit calendar attribut defin granular subset lookup algorithm method gener calendar domain reduc size domain gener graph gener number result shown user type prune reachabl prune preliminari manual prune data rang prune previou discard prune pregener manual prune post gener prune mine associ rule market basket data share measur character itemset problem mine associ rule market basket data recent import research topic area knowledg discoveri databas origin introduc studi extens problem typic examin context discov bui pattern retail sale transact similar data mine applic model studi problem retail store intuit natur clariti retail sale oper larg inventori consist product oper situat locat custom base socio econom divers annual household incom rang low high demograph rang young famili elderli sale manag data mine search associ ru data visual db ddiscov system db discov system research softwar tool knowledg discoveri databas util effici attribut orient induct algorithm base climb tree drop condit method call multi attribut gener gener summari data databas paper present design data visual capabl db discov describ data visual techniqu manag summari potenti interest summari select prototyp system implement altern visual summari result work enabl domain expert quickli effici analyz content databas perspect parallel knowledg discoveri domain gener graph multi attribut gener algorithm attributeori induct relat databas domain gener graph node domain gener graph repres summar domain valu attribut note regener virtual copi attribut dorient induct domain gener graph attribut orient induct summar inform relat databas repeatedli replac specif attribut valu gener concept user defin concept hierarchi show domain gener graph construct multipl concept hierarchi attribut describ graph control gener set attribut present multi attribut gener algorithm attributeori induct domain gener graph base gener test approach algorithm gener combin node domain gener graph individu attribut produc gener relat set attribut rank interesting result gener relat measur base rel entropi varianc experi show measur provid basi analyz summari data relat databas varianc appear rank complex gener relat attribut tupl interest perform analysi regener dbase dynam vote algorithm 