semi dsupervis densiti dbase cluster decompos object dorient class modul agglom cluster techniqu softwar consid live entiti undergo alter lifecycl develop retain good design favor ad featur compli requir meet deadlin reason code complex difficult understand object orient system class larg cohes order identifi problemat case exist approach propos cohes metric metric identifi class low cohes identifi independ concept method requir lot human interpret identifi respect design flaw paper propos class decomposit method agglom cluster algorithm base jaccard distanc class member methodolog identifi concept rank solut impact design qualiti system final method evalu independ design ask comment suggest produc techniqu project design provid feedback abil method identifi concept improv design qualiti system term cohes subspac project cluster experiment evalu analysi find dredund approach propos find correl binari data method focu pair wise correl biologi applic import find correl involv featur set strongli correl featur redund sens correl strong interact featur consid remov featur greatli reduc correl paper explor problem find redund high order correl binari data high order correl formal multi inform gener pairwis mutual inform reduc redund requir subset strongli correl featur subset weakli correl featur subset refer redund interact featur subset nif find nifss computation challeng addit enumer featur combin check subset redund studi properti nifss show properti develop effici algorithm develop set upper lower bound correl incorpor algorithm prune search space simpl effect prune strategi base pair wise mutual inform develop prune search space effici effect approach demonstr extens experi synthet real life dataset pist effici practic index techniqu histor spatio dtempor point data robust project cluster join locat sensor network problem process join queri wireless sensor network focus sensor node process join propos strategi process queri investig perform scenario surprisingli experi show singl strategi perform scenario order avoid potenti high cost fix strategi process queri develop cost base model select join strategi queri hand experi confirm set queri select join strategi base cost model fix strategi queri effect summar multi ddimension data stream histor stream mine adapt process histor spatial rang queri peer dto dpeer sensor network effici composit pattern find monad pattern classif dbase glioma diffus model mri data speedup cluster hierarch rank robust project cluster algorithm learn classif dbase glioma growth model mri data cisr databas system genom dscale comput discoveri regulatori element exploit redund sensor network energi effici process spatiotempor region queri sensor network made autonom devic collect store process share data devic spatiotempor region queri retriev inform interest network queri requir answer subset network node fall queri region network redund sens measur node substitut node degre confid smaller subset node suffici answer queri lower energi cost investig advantag data redund propos techniqu process spatiotempor region queri condit techniqu reduc twenti time energi cost queri process compar typic network flood prolong lifetim sensor network segment brain tumor condit random field support vector machin abstract markov random field mrf popular wellmotiv model medic imag process task segment discrimin random field drf discrimin altern tradition gener mrf tractabl comput restrict simplifi assumpt achiev perform task paper investig tumor segment perform recent variant drf model take advantag power support vector machin svm classif method combin power magnet reson mr preprocess pipelin set align base featur evalu svm mrf type drf classifi segment task relat radiat therapi target plan brain tumor reli contrast agent enhanc result svm base drf offer signific advantag approach analysi spatio dtempor queri process sensor network sensor network emerg technolog provid mean monitor understand phenomena node sensor network typic limit power suppli energi effici process queri network critic issu propos analyt model evalu perform method process histor spatio tempor queri sensor network model understand tradeoff investig method recommend energi effici queri time trajectori split model effici spatio dtempor index paper address problem split trajectori optim purpos effici support spatio tempor rang queri index structur tree minimum bound hyper rectangl trajectori approxim deriv formal cost model estim number os requir evalu spatio tempor rang queri respect queri size arbitrari split trajectori base propos model introduc dynam program algorithm split set trajectori minim number expect disk os respect averag queri size addit develop linear time optim solut problem dynam case trajectori point continu updat experiment evalu confirm effect effici propos split polici embed tree structur methodolog analyz sage librari cancer profil serial analysi gene express sage proven import altern microarrai techniqu global profil mrna popul develop preprocess methodolog address problem analyz sage data due nois caus sequenc error normal methodolog account librari sampl depth miss tag imput methodolog aid analysi poorli sampl sage librari subspac select wilcoxon rank sum test exclud tag similar express level sourc methodolog cluster optic algorithm sage librari deriv cancer normal tissu cell line materi result produc dens cluster repres ovarian cancer cell line brain cancer cell line brain cancer bulk tissu prostat tissu pancreat cancer breast cancer cell line normal brain normal breast bulk tissu ovarian cancer brain cancer cell line cluster close lead investig associ cancer type investig util gene express data classif normal cancer tissu result brain breast cancer librari strong ident allow robust discrimin normal counterpart sage express spatio dtempor databas manag abstract ultim goal intellig travel assist ita fusion develop number technolog devic increas effici transport network increas mass transit ride share ita includ dynam rideshar spatio tempor databas manag wireless commun manag framework ita envis compact portabl devic capabl plan multi modal rout user travel enter desir destin ita formul plan user destin ita make wireless internet technolog send receiv traffic inform transit schedul arrang payment fare global posit system gp track user current posit rout plan mean assess traffic condit en rout spatio tempor databas manag system effici track ita posit real time index term databas system ride share mobil commun transport framework spatio dtempor queri process wireless sensor network wireless sensor network consist node abil measur store process data commun wirelessli node locat wireless rang user issu queri network retriev inform node region applic environment monitor sensor typic limit power suppli energi effici process queri network import issu paper introduc gener framework distribut process spatio tempor queri sensor network main phase rout queri spatial area queri collect process inform node relev queri framework algorithm design independ phase propos algorithm framework phase phase extens experiment evalu studi perform algorithm term energi consumpt vari condit result recommend energi effici solut network spatiotempor queri increment effect data summar dynam hierarch cluster mine inform pattern larg dynam chang databas pose numer interest challeng data summar data bubbl propos compress larg static databas repres point suitabl subsequ effect hierarch cluster analysi real world applic databas dynam chang due frequent insert delet possibl chang data distribut cluster structur time complet reappli data summar cluster algorithm detect cluster structur updat uncov data pattern delet insert prohibit expens larg fast chang databas paper propos scheme maintain data bubbl increment increment data bubbl high qualiti hierarch cluster quickli point time scheme qualiti measur increment data bubbl identifi data bubbl compress underli data point insert delet data bubbl built effici split merg oper extens experiment evalu show increment data bubbl provid significantli faster data summar complet build data bubbl number insert delet effect preserv case improv qualiti data summar automat extract cluster hierarch cluster represent hierarch cluster algorithm typic effect detect true cluster structur data set partit algorithm effici index high dimension normal histogram paper address problem index high dimension normal histogram data dimension featur vector sum represent multimedia object order facilit similar queri process analyz properti induc constraint hold gener multi dimension space design split polici show perform similar queri normal histogram data signi cantli improv exploit properti simpl index framework process nearest neighbor queri time faster sr tree time faster tree automat extract cluster hierarch cluster represent hierarch cluster algorithm typic effect detect true cluster structur data set partit algorithm data bubbl dvector data speed dup hierarch cluster arbitrari metric space speed cluster algorithm data summar method propos summar data set comput suitabl repres object cluster algorithm appli repres cluster structur data set deriv base result repres previou method limit applic domain gener base suffici statist linear sum set point assum data vector space hand import applic data metric vector space distanc object exploit construct effect data summar paper develop data summar method base distanc inform appli directli vector data extens perform evalu show method effect find hierarch cluster structur vector data small number data summar result larg reduct runtim trade cluster qualiti focus dcitat improv retriev relat page web analysi sigmod dauthorship graph hierarch cluster analysi sage data cancer profil paper present method cluster sage serial analysi gene express data detect similar dissimilar type cancer subcellular level data extrem high dimension due method measur error miss valu data challeng cluster algorithm introduc special pre process techniqu reduc error restor miss data techniqu tailor process gener data make conserv present subspac select techniqu identifi relev subset attribut gene wilcoxon test gener techniqu appli select subspac purpos cluster high level categori interest data cancer noncancer final discuss result applic cluster algorithm optic sage data preprocess step data bubbl qualiti preserv perform boost hierarch cluster paper investig scale hierarch cluster method optic extrem larg databas util data compress method birch random sampl propos step procedur compress data suitabl repres object appli hierarch cluster algorithm object recov cluster structur data set base result compress data kei issu approach design compress data item hierarch cluster algorithm appli inform infer cluster structur origin data set step crucial result hierarch cluster algorithm appli naiv random sampl cluster featur cf gener birch deterior rapidli higher compress rate due kei problem identifi solv problem propos effici post process step concept data bubbl special kind compress data item appli optic data bubbl recov accur approxim cluster structur larg data set high compress rate comprehens perform qualiti evalu show trade qualiti cluster result great increas perform multipl similar queri basic dbm oper mine metric databas knowledg discoveri databas techniken und anwendungen knowledg discoveri databas techniken und anwendungen effici support multipl similar queri mine metric databas metric databas databas metric distanc function defin pair databas object databas similar queri form rang queri nearest neighbor queri import queri tradit queri process singl queri issu independ user data mine applic databas typic explor iter similar queri answer previou similar queri paper introduc gener scheme data mine algorithm develop method transform algorithm multipl similar queri set queri issu simultan investig orthogon approach reduc cost cpu cost speed process multipl similar queri propos techniqu appli type similar queri implement base index sequenti scan paral independ quantiz index compress techniqu high ddimension data space major approach propos effici process queri databas speed search index structur speed search oper compress databas signatur file approach limit index techniqu ineffici extrem configur high dimension space simpl scan cheaper index base search compress techniqu effici situat propos combin techniqu search nearest neighbor high dimension space purpos develop compress index call iq tree level structur level regular flat directori consist minimum bound box level data point compress represent level actual data overcom engin challeng construct effect index structur type fast hierarch cluster base compress data optic scale cluster algorithm squash data intellig compress techniqu cluster compress data record compress data record produc birch algorithm typic consist suffici statist form number point vector sum squar sum point directli speed mean type cluster algorithm obviou hierarch cluster algorithm appli hierarch cluster algorithm center compress subclust produc weak result reason hierarch cluster algorithm base distanc data point interpretaion result reli heavili correct graphic represent distanc paper introduc method suffici statist lof identifi densiti dbase local outlier kdd applic detect crimin activ commerc find rare instanc outlier interest find common pattern exist work outlier detect outlier binari properti paper contend scenario meaning assign object degre outlier degre call local outlier factor lof object local degre depend isol object respect surround neighborhood give detail formal analysi show lof enjoi desir properti realworld dataset demonstr lof find outlier meaning identifi exist approach final care perform evalu algorithm confirm show approach find local outlier practic spatial data mine databas primit algorithm effici dbm support spatial data mine algorithm heavili depend effici process neighborhood relat neighbor object investig singl run typic algorithm provid gener concept neighborhood relat effici implement concept tight integr spatial data mine algorithm spatial databas manag system speed develop execut spatial data mine algorithm paper defin neighborhood graph path small set databas primit manipul show typic spatial data mine algorithm support propos basic oper find signific spatial pattern class path lead start object relev discuss filter allow neighborhood path significantli reduc search space spatial data mini databas primit spatial data mine spatial data mine algorithm heavili depend effici process neighborhood relat neighbor object investig singl run typic algorithm provid gener concept neighborhood relat effici implement concept tight integr spatial data mine algorithm spatial databas manag system speed develop execut spatial data mine algorithm paper defin neighborhood graph path small set databas primit manipul introduc neighborhood indic speed process databas primit implement databas primit top commerci spatial databas manag system effect effici propos approach evalu analyt cost model extens experiment studi geographi knowledg discoveri spatial databas number size spatial databas geograph medic databas rapidli grow larg amount data obtain satellit imag comput tomographi scientif equip knowledg discoveri databas kdd process discov valid potenti pattern larg databas typic task knowledg discoveri spatial databas includ cluster character trend detect major differ knowledg discoveri relat databas spatial databas attribut neighbor object interest influenc object spatial knowledg discoveri algorithm heavili depend effici process neighborhood relat neighbor object investig singl run typic algorithm provid gener concept neighborhood relat eff knowledg discoveri spatial databas number size spatial databas geograph medic databas rapidli grow larg amount data obtain satellit imag comput tomographi scientif equip knowledg discoveri databas kdd process discov valid potenti pattern larg databas typic task knowledg discoveri spatial databas includ cluster character trend detect major differ knowledg discoveri relat databas spatial databas attribut neighbor object interest influenc object spatial knowledg discoveri algorithm heavili depend effici process neighborhood relat neighbor object investig singl run typic algorithm provid gener concept neighborhood relat eff knowledg discoveri spatial databas number size spatial databas geograph medic databas rapidli grow larg amount data obtain satellit imag comput tomographi scientif equip knowledg discoveri databas kdd process discov valid potenti pattern larg databas typic task knowledg discoveri spatial databas includ cluster character trend detect major differ knowledg discoveri relat databas spatial databas attribut neighbor object interest influenc object spatial knowledg discoveri algorithm heavili depend effici process neighborhood relat neighbor object investig singl run typic algorithm provid gener concept neighborhood relat eff optic dof identifi local outlier kdd applic find outlier rare event interest find common case detect crimin activ commerc outlier binari properti optic dof identifi local outlier kdd applic find outlier rare event interest find common case detect crimin activ commerc outlier binari properti gener densiti dbase cluster spatial data mine abstract qualit plan dbase simul intern crise distribut dbase cluster algorithm mine larg spatial databas problem detect cluster point belong spatial point process aris applic paper introduc cluster algorithm dbclasd distribut base cluster larg spatial databas discov cluster type result experi demonstr dbclasd contrari partit algorithm claran discov cluster arbitrari shape dbclasd requir input paramet contrast cluster algorithm dbscan requir input paramet difficult provid larg databas term effici dbclasd claran dbscan close dbscan effici dbclasd larg spatial databas attract algorithm character trend detect spatial databas number size spatial databas geomarket traffic control environment studi rapidli grow result increas spatial data mine paper present algorithm spatial character spatial trend analysi spatial character import class membership databas object determin spatial attribut attribut object neighborhood spatial trend analysi pattern chang spatial attribut neighborhood databas object determin present algorithm task algorithm implement gener framework spatial data mine provid small set databas primit top commerci spatial databas manag system perform evalu real geograph databas demonstr effect propos algorithm show algorithm combin discov interest spatial knowledg increment cluster mine data wareh environ data warehous provid great deal opportun perform data mine task classif cluster typic updat collect appli data warehous period batch mode night pattern deriv warehous data mine algorithm updat due larg size databas highli desir perform updat increment paper present increment cluster algorithm algorithm base cluster algorithm dbscan applic databas data metric space spatial databas www log databas due densiti base natur dbscan insert delet object affect current cluster neighborhood object effici algorithm increment insert delet exist cluster base formal definit cluster proven increment algorithm yield result dbscan perform evalu incrementaldbscan spatial databas www log databas present demonstr effici propos algorithm incrementaldbscan yield signific speed factor dbscan larg number daili updat data warehous densiti dbase cluster spatial databas algorithm gdbscan applic cluster mine larg spatial databas densiti dconnect set applic trend detect spatial databas cluster algorithm propos class identif spatial databas earth observ databas effect algorithm dbscan limit fulli exploit rich type data contain spatial databas paper introduc concept densiti connect set present significantli gener version dbscan major properti algorithm symmetr predic defin neighborhood object allow natur definit case spatial extend object polygon cardin function set neighbor object account spatial attribut object mean assign applic specif weight densiti connect set basi discov trend spatial databas defin trend spatial databas show appli gener dbscan algorithm task discov knowledg demonstr practic impact approach perform experi geograph inform system bavaria repres broad class spatial databas spatial data mine databas approach knowledg discoveri databas kdd import task spatial databas number size databas rapidli grow paper introduc set basic oper support spatial databas system sdb express algorithm kdd sdb purpos introduc concept neighborhood graph path small set oper manipul argu oper suffici kdd algorithm spatial neighborhood relat present implement typic spatial kdd algorithm base propos oper effici support oper larg neighborhood graph larg set neighborhood path sdb discuss neighborhood indic introduc materi select neighborhood graph order speed process propos oper densiti dbase algorithm discov cluster larg spatial databas nois cluster algorithm attract task class identif spatial databas applic larg spatial databas rise requir cluster algorithm minim requir domain knowledg determin input paramet discoveri cluster arbitrari shape good effici larg databas cluster algorithm offer solut combin requir paper present cluster algorithm dbscan reli densiti base notion cluster design discov cluster arbitrari shape dbscan requir input paramet support user determin perform experiment evalu effect effici dbscan synthet data real data sequoia benchmark result experi demonstr dbscan significantli effect discov cluster arbitrari shape algorithm clar an dbscan outperform claran factor term effici 