formal concept sampl count threshold dfree local pattern mine privaci dpreserv data dmine paper address issu privaci preserv data mine specif scenario parti own confidenti databas run data mine algorithm union databas reveal unnecessari inform work motiv protect privileg inform enabl research purpos dredund subgroup discoveri closur system subgroup discoveri numer domain subgroup discoveri numer domain approxim number frequent set dens data tight optimist estim fast subgroup discoveri abstract subgroup discoveri task find subgroup popul exhibit distribut unusu high gener due monoton evalu function standard prune techniqu subgroup discoveri requir optimist estim techniqu optimist estim prune consid extrem simpl case binari target attribut attempt made move suboptim heurist optimist estim paper show optimist estim prune develop sound highli effect prune approach subgroup discoveri base precis definit optim show previou estim tight special case present tight optimist estim popular binari multi class qualiti function present famili increasingli effici approxim optim function show empir experi newli propos optimist estim lead speed order magnitud compar previou approach random approach approxim number frequent set procur fraud discoveri similar measur learn abstract paper describ approach detect hint procur fraud develop context european union project fraud prevent procur fraud special kind fraud occur employe cheat employ execut trigger bogu payment approach present base idea learn similar measur compar employe payrol stand data record creditor record order detect creditor suspici similar employe end combin simpl similar measur address similar spacial similar weight scheme weight similar function learn user input pair payrol creditor data record similar kei word similar measur learn fraud detect health care procur fraud discoveri similar measur learn abstract paper describ approach detect hint procur fraud develop context european union project fraud prevent procur fraud special kind fraud occur employe cheat employ execut trigger bogu payment approach present base idea learn similar measur compar employe payrol stand data record creditor record order detect creditor suspici similar employe end combin simpl similar measur address similar spacial similar weight scheme weight similar function learn user input pair payrol creditor data record similar kei word similar measur learn fraud detect health care probabilist hybrid action model predict concurr percept ddriven robot behavior articl develop probabilist hybrid action model pham realist causal model predict behavior gener modern percept driven robot plan pham repres aspect robot behavior repres action model ai plan tempor structur continu control process nondeterminist effect mode interfer ccgolog logic languag deal continu chang dline execut cc dgolog plan previous plan languag cc golog introduc purpos event driven behavior typic found robot control cc golog usabl project outcom plan unclear execut plan line robot paper provid execut model ccgolog addit show interleav execut kind time bound project belief updat pgolog framework high level control oper robot dynam uncertain domain concern reason task deal effect noisi sensor effector project effect candid plan updat belief line execut plan paper show pgolog framework origin form account project high level plan extend reason robot belief evolv line execut plan pgolog extens high level program languag golog specif probabilist belief state world represent sensor effector uncertain probabilist outcom applic belief updat introduc belief base program golog style program test appeal agent belief execut time integr plan dbase control autonom robot human environ author extend rhino robot ad mean plan base high level control plan transform enhanc probabilist reason capabl result autonom robot capabl accomplish ing prolong complex dynam chang task real world cc dgolog realist logic dbase robot control probabilist hybrid action model predict concurr percept ddriven robot behavior articl develop probabilist hybrid action model pham realist causal model predict behavior gener modern percept driven robot plan pham repres aspect robot behavior repres action model ai plan tempor structur continu control process nondeterminist effect mode interfer turn high dlevel plan robot program uncertain domain action robot lift object thought low level process uncertain outcom highlevel robot plan descript task combin process involv nondetermin order increas plan gener situat robot turn plan execut program establish form project satisfi goal probabl paper show achiev logic framework low level process model program pgolog probabilist variant action languag golog high level plan ordinari golog program project name low level process replac pgolog definit introduct action robot lift object thought low level process uncertain ou cc dgolog realist logic dbase robot control realist logic dbase robot control golog framework paper show issu dealt golog frameworkbi propos extens languag probabilist abstract paper present bayesian method construct probabilist network databas focu construct bayesian belief network potenti applic includ comput assist hypothesi test autom scientif discoveri autom construct probabilist expert system extend basic method handl miss data hidden latent variabl show perform probabilist infer averag infer multipl belief network result present preliminari evalu algorithm construct belief network databas case final relat method paper previou work discuss open problem probabilist abstract paper present bayesian method construct probabilist network databas focu construct bayesian belief network potenti applic includ comput assist hypothesi test autom scientif discoveri autom construct probabilist expert system extend basic method handl miss data hidden latent variabl show perform probabilist infer averag infer multipl belief network result present preliminari evalu algorithm construct belief network databas case final relat method paper previou work discuss open problem 