frontier appli data mine visual tempor cluster rel densiti dorgan map visual tempor cluster rel densiti dorgan map redsom rel densiti visual tempor cluster structur dorgan map exploratori hot spot profil analysi interact visual drill ddown dorgan map mine unexpect tempor associ applic detect advers drug reaction exploratori multilevel hot spot analysi australian taxat offic case studi survei open sourc data mine system data mine analyt crc construct innov deleg contain claus neighborhood densiti method select initi cluster center dmean cluster data mine theori methodolog techniqu applic neighborhood densiti method select initi cluster center dmean cluster identifi risk group colorect cancer deliveri framework health data mine analyt health explor tool develop csiro doha deliv web servic type data mine analyt facil web interfac provid desktop access sophist analys larg data collect tool user access larg transact dataset creat profil select patient patient profil window event sequenc data analyz user chosen data mine tool result analysi visual form knowledg represent method initi implement tool focus applic advers drug reaction explor tool embed data mine tool broad area health data analysi mine risk pattern medic data paper discuss problem find risk pattern medic data defin risk pattern statist metric rel risk wide epidemiolog research characteris problem mine risk pattern optim rule discoveri problem studi anti monoton properti mine optim risk pattern set present algorithm make properti risk pattern discoveri method appli real world data set find pattern allerg event ac inhibitor algorithm gener result medic research categori subject descriptor repres associ classif rule mine health data abstract associ classif algorithm develop explor advers drug reaction larg medic transact dataset unbalanc class rule discov alert medic practition prescrib drug categori patient potenti advers effect assess rule surviv chart propos kind probabl tree present repres risk advers drug reaction categori patient term risk ratio familiar medic practition approach show risk ratio rule condit appli present risk singl risk factor part rule identifi cohort patient subpopul probabl tree present risk specif advers drug reaction prescrib repres associ classif rule mine health data abstract associ classif algorithm develop explor advers drug reaction larg medic transact dataset unbalanc class rule discov alert medic practition prescrib drug categori patient potenti advers effect assess rule surviv chart propos kind probabl tree present repres risk advers drug reaction categori patient term risk ratio familiar medic practition approach show risk ratio rule condit appli present risk singl risk factor part rule identifi cohort patient subpopul probabl tree present risk specif advers drug reaction prescrib explor advers drug reaction cluster event sequenc conceptu mine larg administr health data tempor sequenc associ rare event conceptu mine larg administr health data dline unsupervis outlier detect finit mixtur discount learn algorithm outlier detect fundament issu data mine speci calli fraud detect network intrus detect network monitor smartsift abbrevi ss outlier detect engin adrress problem viewpoint statist learn theori paper theoret basi ss empir demonstr effect ss detect outlier onlin process line unsupervis learn probabilist model finit mixtur model inform sourc time datum input ss emploi line discount learn algorithm learn probabilist model score datum base learn model high score indic high possibl statist outlier featur ss adapt stationari sourc data score clear statist inform theoret mean computation inexpens handl categor continu variabl experiment applic network intrus detect show ss identifi data high score correspond attack low comput cost experiment applic identifi number meaning rare case actual health insur patholog data australia health insur commiss associ rule discoveri unbalanc class distribut method find associ rule larg data gener associ rule discoveri method find rule includ lot uninterest rule perform algorithm deterior minimum support low fail find interest rule support low case significantli unbalanc class paper present algorithm find associ rule base set interesting criteria algorithm appli real world health data set successfulli identifi group patient high risk advers reaction drug statist guid method select featur develop initi result shown propos algorithm find interest pattern data set unbalanc class distribut perform loss mine data stream abstract mine data stream rais number research challeng data mine commun challeng includ limit comput resourc mine stream data mobil devic limit resourc due continu data stream algorithm pass incom data record articl algorithm output granular aog approach mine data stream discuss aog adapt approach cope challeng inher featur data stream show result aog base cluster resourc constrain environ estim episod care link medic claim data australia extens administr health data collect commonwealth state agenc health data quickli ectiv data convert knowledg independ quantiti qualiti uniqu clean link administr health dataset address problem empir defin episod care episod care time interv medic servic relat medic situat paper medic situat hospit admiss medic servic interest patholog test diagnost imag noninvas investig procedur perform hospit admiss hospit admiss task view detect signal time seri relat hospit admiss distinct background nois medic care approach ensembl panel expert paradigm implement multipl agent altern predict model independ estim interv combin good interv estim vote scheme result studi commonwealth depart health outlier detect replic neural network problem find outlier larg multivari databas outlier detect appli data cleans process data mine identifi problem data fraud detect group outlier interest mine tempor pattern health care data compar studi rnn outlier detect data mine propos replic neural network rnn outlier detect algorithm compar rnn outlier detect method publicli statist dataset gener small data mine dataset gener larger gener real data smaller dataset provid insight rel strength weak rnn compar method larger dataset test scalabl practic applic paper develop methodolog compar outlier detector perform benchmark outlier detect method assess knowledg discoveri data mine pakdd advanc artifici intellig pricai workshop reader featur select tempor health record paper altern featur vector represent patient health record longitudin tempor irregular charact patient episod histori integr part health record challeng appli data mine techniqu tempor data mine hidden markov dlocal polynomi model tempor data mine multilevel dlocal polynomin model dline unsupervis outlier detect finit mixtur discount learn algorithm outlier detect fundament issu data mine speci calli fraud detect network intrus detect network monitor smartsift abbrevi ss outlier detect engin adrress problem viewpoint statist learn theori paper theoret basi ss empir demonstr effect ss detect outlier onlin process line unsupervis learn probabilist model finit mixtur model inform sourc time datum input ss emploi line discount learn algorithm learn probabilist model score datum base learn model high score indic high possibl statist outlier featur ss adapt stationari sourc data score clear statist inform theoret mean computation inexpens handl categor continu variabl experiment applic network intrus detect show ss identifi data high score correspond attack low comput cost experiment applic identifi number meaning rare case actual health insur patholog data australia health insur commiss artifici intellig industri industri applic larg central softwar system effect distribut network simpler computer agent compet effect todai market manufactur design implement reconfigur resiz maintain manufactur facil rapidli inexpens modern manufactur depend heavili comput system requir appli manufactur control softwar easili satisfi small modul larg monolith system paper review industri distribut artifici intellig dai give special attent system manufactur schedul control describ taxonomi system case studi advanc research applic actual industri instal identifi step deploi technolog broadli mine taxat data parallel bmar parallel version friedman multivari adapt regress spline mar algorithm discuss partit data processor parallel comput system achiev good parallel ecienc truncat power basi function origin mar method bmar utilis spline improv numer stabil reduc comput cost procedur addit coecient basi function bmar model provid quickli access inform local behaviour function algorithm time complex proport number data record method mean detect area space featur characteris interest pattern respons valu appli search class incorrect tax return multipl predictor variabl featur integr deliveri larg dscale data mine acsi data mine project evolutionari hot spot data mine architectur explor interest discoveri mine knowledg mine hot spot methodolog mine larg real world databas databas grow size complex task ad wealth data difficult data mine emerg technolog add enorm databas find import snippet nugget knowledg larg train set extrem larg collect nugget extract lead fool gold fossick real gold attent direct problem focu preciou nugget paper present hot spot methodolog adopt multi strategi interact approach focu import nugget methodolog perform data mine explor result model find import nugget contain approach demonstr insur fraud applic introduct rapid increas databas dramat increas storag capac perform man pepnet parallel evolutionari program construct artifici neural network paper present descript evolutionari artifici neural network algorithm ep net extens take advantag high perform comput environ pep net parallel epnet implement form parallel paper describ parallel experiment studi shown promis result time predict perform templat spatial reason respons geograph inform system 