analyz chang spatial data util polygon model correspond cluster approach cluster multipl relat spatial dataset framework multi dobject cluster applic dlocat mine abstract goal multi object cluster moc decompos dataset similar group maxim multipl object parallel paper provid methodolog architectur algorithm base larg set object deriv interest cluster object propos architectur reli cluster algorithm support plug fit function multi run cluster cluster algorithm run multipl time maxim subset object captur compound fit function moc search engin type capabl user enabl queri larg set cluster respect object threshold evalu propos moc framework case studi center spatial locat mine goal identifi region high level arsen concentr locat high concentr chemic texa water suppli keyword multi object cluster knowledg discoveri spatial data mine locat mine cluster plug fit function architectur algorithm multi drun cluster abstract paper address main challeng cluster requir extens human effort select paramet arbitrari cluster algorithm identifi altern cluster propos architectur concret system mr clever multi run cluster integr activ learn cluster algorithm kei hypothesi work cluster result obtain combin cluster origin multipl run cluster algorithm defin state repres paramet set cluster algorithm propos architectur activ learn state util function util paramet set assess base cluster run time qualiti novelti obtain cluster util function plai import role guid cluster algorithm seek solut cluster novelti measur introduc purpos final contribut cluster summar algorithm assembl final cluster combin high qualiti cluster origin multipl run merit propos system gener conjunct cluster algorithm reduc human effort select paramet compar cluster result assembl cluster result evalu propos system conjunct repres base cluster algorithm clever challeng data mine task involv earthquak dataset obtain result demonstr comparison singl run cluster multi run cluster discov solut higher qualiti reg region regress framework geo dreferenc dataset tradit regress analysi deriv global relationship variabl neglect spatial variat variabl lack abil systemat discov region relationship build model region knowledg obtain higher predict accuraci relationship spatial dataset region great region regress method deriv region regress function reflect spatial characterist region paper propos region regress framework discov interest region show strong region relationship depend independ variabl build predict model region regress function region interest region identifi run repres base cluster algorithm maxim extern plug fit function work propos fit function squar base fit function aic base fit function handl overfit evalu framework case studi identifi arsen contamin texa water well boston hous dataset determin spatial vari effect hous properti hous price demonstr framework effect identifi interest region build predict system reli region model region pattern discoveri geo dreferenc dataset pca abstract exist data mine techniqu focu find global pattern lack abil systemat discov region pattern relationship spatial dataset region great extract region knowledg spatial dataset paper propos framework discov interest region character strong region correl relationship attribut method analyz differ similar region framework emploi twophas approach discov region emploi cluster algorithm maxim pca base fit function appli post process techniqu explain underli region structur correl pattern addition similar measur assess structur similar region base correl set introduc evalu framework case studi center find correl arsen pollut factor water well demonstr framework effect identifi region correl pattern chang analysi spatial data combin contour algorithm supervis densiti function abstract detect spatial dataset import field paper introduc methodolog chang analysi spatial dataset combin contour algorithm supervis densiti estim techniqu methodolog user defin criteria featur interest identifi featur dataset chang analysi perform compar interest region deriv contour cluster cluster algorithm call dcontour introduc purpos comput contour polygon describ boundari supervis densiti function densiti threshold relationship data analyz reli polygon oper evalu methodolog case studi analyz earthquak pattern chang analysi spatial dataset interesting comparison acm sigspati address issu relat acquisit manag process spatial relat inform focu algorithm geometr visual consider scope includ limit geograph inform system current acm sigspati offic find region dlocat pattern set continu variabl spatial dataset discov control factor geospati variabl effici mean determin factor control spatial distribut environment class variabl signific interest earth scienc paper present method autom discoveri control factor mine emerg pattern databas construct fusion explanatori dataset introduc definit pattern support account spatial charact data systemat evalu effect techniqu real world applic pertain densiti veget cover experiment result show method successfulli identifi control factor presenc high veget cover region discoveri spatial dataset abstract paper present region discoveri framework gear find scientif interest place spatial dataset view region discoveri cluster problem extern fit function maxim framework adapt repres cluster algorithm exemplifi prototyp base gridbas densiti base agglom cluster algorithm systemat evalu algorithm real world case studi task find featur base hotspot extrem densiti deep ic shallow ic locat mar result reveal densiti base algorithm outperform algorithm discov region higher interesting grid base algorithm provid accept solut quickli agglom cluster algorithm perform identifi larger region arbitrari shape result region mar shallow deep ground ic locat suggest deposit geolog time distanc function learn supervis similar assess mosaic proxim graph approach agglom cluster abstract repres base cluster algorithm popular due rel high speed sound theoret foundat hand cluster obtain limit convex shape cluster result highli sensit initi paper agglom cluster algorithm call mosaic propos greedili merg neighbor cluster maxim fit function mosaic gabriel graph determin cluster neighbor approxim convex shape union small cluster comput repres base cluster algorithm experiment result show techniqu lead cluster higher qualiti compar run repres cluster algorithm standalon suitabl fit function mosaic detect arbitrari shape cluster addit mosaic capabl deal high dimension data keyword post process hybrid cluster find cluster arbitrari shape agglom cluster proxim graph cluster supervis densiti estim techniqu applic spatial data mine region associ rule scope special challeng spatial data mine inform distribut uniformli spatial data set discoveri region knowledg fundament import region pattern frequent fail discov due insuffici global confid support tradit associ rule mine region associ rule definit hold subspac global space challeng evalu impact region associ rule paper center region associ rule scope introduc reward base region discoveri framework emploi cluster find place region associ rule valid evalu approach real world case studi discov arsen risk zone texa water suppli experiment result valid domain expert compar publish result arsen contamin discoveri interest region spatial data set supervis cluster abstract discoveri interest region spatial dataset import data mine task interest identifi disjoint contigu region unusu respect distribut class region unusu low high number instanc class paper center discuss techniqu methodolog algorithm discov region measur interesting supervis cluster framework introduc purpos supervis cluster algorithm propos paper agglom hierarch supervis cluster name scah agglom grid base cluster method name schg lastli algorithm name scmrg search multi resolut grid structur top interest region final experiment result appli propos framework algorithm problem identifi hotspot spatial dataset discuss cluster learn distanc function supervis similar assess adapt cluster obtain cluster feedback past experi adapt cluster extern feedback improv cluster qualiti past experi serv speed execut time adapt cluster environ propos learn learn reward valu success data cluster adapt cluster support reus cluster memor work past capabl explor multipl path parallel search good cluster case studi appli adapt cluster instanc base learn reli distanc function modif approach distanc function adapt scheme extern feedback propos compar distanc function learn approach experiment result adapt cluster lead signific improv instanc base learn techniqu nearest neighbor classifi product instanc base learn techniqu introduc classifi exampl sole cluster repres techniqu show high promis experiment evalu content dbase imag retriev multi dagent meta dlearn framework object gener purpos content base imag retriev system find imag databas match extern measur relev user follow inconsist relev measur process queri task specif manner shown effect approach view special imag retriev algorithm agent propos gener purpos imag retriev system multi agent meta learn framework framework adapt distanc function defin imag distanc weight imag queri identifi cluster algorithm produc similar solut similar problem experi compar approach tradit inform retriev algorithm result show framework averag relev score supervis cluster enhanc classifi abstract paper center data mine techniqu term supervis cluster unlik tradit cluster supervis cluster appli classifi exampl goal identifi class uniform cluster high probabl densiti paper focus data mine techniqu gener classif techniqu benefit knowledg obtain supervis cluster discuss nearest neighbor classifi construct knowledg gener supervis cluster provid experiment evid effici accur tradit nearest neighbor classifi final demonstr supervis cluster enhanc simpl classifi cluster learn distanc function supervis similar assess dmedoid dstyle cluster algorithm supervis summari gener paper center discuss medoid style cluster algorithm supervis summari gener task requir cluster techniqu identifi class uniform cluster paper investig cluster techniqu term supervis cluster work focus gener medoid style cluster algorithm investig supervis cluster algorithm sridhcr singl repres insert delet hill climb restart spam variat pam solut qualiti run time algorithm tradit cluster algorithm pam evalu benchmark consist data set experi show supervis cluster algorithm enhanc class puriti tradit cluster algorithm pam sridhcr find solut spam kei word supervis summari gener cluster classifi exampl medoid cluster algorithm data mine systemat databas summari gener distribut queri discoveri system repres dbase cluster nearest neighbor dataset edit goal dataset edit instanc base learn remov object train set order increas accuraci classifi wilson edit remov train exampl misclassifi nearest neighbor classifi smooth shape result decis boundari paper revolv repres base cluster algorithm nearest neighbor dataset edit term approach supervis cluster edit main idea replac dataset set cluster prototyp cluster approach call supervis cluster introduc purpos empir evalu uci dataset show wilson supervis cluster edit improv accuraci dataset test supervis cluster edit achiev time higher compress rate wilson edit obtain significantli high accuraci dataset test supervis cluster algorithm benefit paper center data mine techniqu term supervis cluster unlik tradit cluster supervis cluster assum exampl classifi goal identifi class uniform cluster high probabl densiti repres base algorithm supervis cluster introduc greedi algorithm random restart name sridhcr seek solut insert remov singl object current solut spam variat cluster algorithm pam evolutionari comput algorithm name scec fast medoid base top split algorithm name td algorithm evalu benchmark consist uci machin learn data set gener greedi algorithm spam sridhcr td perform supervis cluster termin prematur briefli describ applic supervis cluster class decomposit cluster framework low dvarianc classifi paper propos pre process step classif appli cluster algorithm train set discov local pattern attribut input space demonstr knowledg exploit enhanc predict accuraci simpl classifi focu classifi character high bia low varianc linear classifi classifi experi difficulti delin class boundari input space class distribut complex wai decompos class cluster make class distribut easier approxim viabl reduc bia limit growth varianc experiment result real world domain show advantag predict accuraci cluster preprocess step classif fast decis tree learn algorithm microarrai data collect heurist search base factor tool emperor cheap legal secur cryptographi web support polyploidi genet algorithm domin vector offset invari crossov tool discov cycl length period function support polyploidi genet algorithm domin vector recurr select improv ga perform recurr select improv ga perform deriv queri result genet program enhanc divers genet algorithm learn environ classfic task adapt brows system base rule object dorient databas evalu enhanc bayesian rule dset genet algorithm learn environ classif task paper describ induct learn environ call delvaux classif task learn prospector style bayesian rule set exampl genet algorithm approach learn bayesian rule set popul consist set rule set gener offspr exchang rule permit fitter rule set produc offspr higher probabl reward punish mechan introduc evalu perform bayesian rule rule set purpos fuzzi techniqu evalu good rule rule set provid mutat oper introduc evalu inform replac bad rule higher probabl good rule mutat occur empir result evalu present reward punish techniqu present final compar learn environ learn fuzzi rule reli decis make evid combin classic classifi system reli fuzzi data driven rule bucket brigad algorithm kei word evolutionari comput learn adapt system genet algorithm classifi system bayesian rule base system induct learn approxim reason knowledg discoveri bucket brigad algorithm automat rule debug learn bayesian classif rule genet algorithm express enforc dynam integr constraint integr set rule data object dorient environ rule dbase consist enforc knowledg dbase system rule extend suggest fault data ddriven rule dbase system activ pattern control rule integr data ddriven command ddriven program integr forward dchain rule oper perman knowledg base tangui integr databas rule dbase object dorient paradigm integr variabl oper rule dbase forward chain system methodolog design transform conceptu schema formal semant infer rule conceptu data model integr rule knowledg base manag system dali knowledg base manag system acquisit terminolog knowledg databas design techniqu natur languag requir good data base definit data base design methodolog problem solv hyper plan 