supervis identif link concept mention domain dspecif ontolog opinion digger unsupervis opinion miner unstructur product review fvc featur dvector dbase classif xml dissemin matrix factor techniqu trust propag recommend social network semi dsupervis learn posit unlabel data entropi data mine approach xml dissemin psortb improv protein subcellular local predict refin local subcategori predict capabl prokaryot infer cancer subnetwork marker densiti dconstrain biclust collabor mine multipl social network data crimin group discoveri multi drelat approach spatial classif trustwalk random walk model combin trust dbase item dbase recommend dens graphlet statist protein interact random network trust network improv top dn recommend feedbacktrust feedback effect trust dbase recommend system mine cohes pattern graph featur vector quantifi system evolutionari color code confid dscore ppi network abstract current major challeng system biologi comput statist biomolecular network motif reveal signific system differ organ extend color code techniqu weight edg network appli ppi network edg weight probabilist confid score provid string databas substanti improv previous studi heavili noisi binari edg weight data studi comput expect number occurr induc subtre vertic previous report differ unicellular multicellular organ reveal major differ prokaryot unicellular eukaryot establish time statist sound data basi evolutionari distanc monitor term elev system arrang densiti dbase cluster abstract densiti base cluster method identif arbitrari necessarili convex region data point dens popul number cluster cluster defin connect region exce densiti threshold paper introduc notion local scale densiti base cluster determin densiti threshold base local statist data local maxima densiti discov nearest neighbor densiti estim center potenti cluster cluster grown densiti fall pre ratio center point densiti result cluster techniqu identifi cluster arbitrari shape noisi background signific densiti gradient focu paper autom process cluster make local densiti inform arbitrarili size shape locat number cluster perform algorithm promis demonstr number synthet dataset imag wide rang paramet slider maximum probabl inform align short sequenc read snp detect kiwi scalabl subspac cluster algorithm gene express analysi join bay net type bay net relat data real world data maintain relat format tabl store inform entiti link relationship structur schema databas essenti logic languag variabl rang individu entiti predic relationship attribut work combin graphic structur bay net logic structur relat databas achiev knowledg discoveri databas introduc join bay net type bay net repres learn class level depend attribut tabl tabl depend import polici make strateg plan focus class level depend bring advantag term simplic model tractabl infer learn usual bay net graphic structur support effici infer reason show appli standard bay net infer algorithm learn model fast accur probabl estim queri involv attribut relationship multipl tabl robust project cluster joint cluster analysi attribut data relationship data connect dcenter problem multi drelat skylin oper exist work skylin queri extens decis support recommend system focus effici issu singl tabl data retriev user target skylin store multipl tabl requir perform join oper tabl result cost comput skylin join tabl increas dramat due potenti increas cardin dimension paper systemat studi skylin oper multi relat databas propos solut aim seamlessli integr state art join method skylin comput experi demonstr propos method effici show promis applic extend skylin oper typic databas oper join aggreg effici increment constrain cluster cluster constraint emerg area data mine research work assum constraint larg batch paper explor situat constraint increment user cluster provid posit neg feedback constraint critiqu cluster solut problem effici updat cluster satisfi constraint cluster entir data set show problem increment cluster constraint np hard gener identifi suffici condit lead effici solvabl version translat set rule type constraint ad constraint set properti maintain demonstr approach effici cluster entir data set advantag minimum consist subset cover problem applic data mine paper introduc studi minimum consist subset cover mcsc problem finit ground set constraint find minimum number consist subset cover subset consist satisfi mcsc problem gener tradit set cover problem minimum cliqu partit dual problem graph color instanc practic data mine problem area rule learn cluster frequent pattern mine formul mcsc instanc discuss minimum rule set problem minim model complex decis rule convers cluster problem minim number cluster satisfi distanc constraint show mcsc problem find applic frequent pattern summar mcsc formul propos graphbas gener algorithm cag directli applic cag start construct maxim optim partial solut perform driven specif gener search dynam maintain bipartit assign graph simultan learn set consist subset small cardin cover ground set experi benchmark dataset show cag achiev good result compar exist popular heurist constraint ddriven cluster cluster method data driven driven data driven method intend discov true structur underli data driven method aim organ true structur meet applic requir driven constrain cluster find action cluster applic energi awar sensor network privaci preserv market segment exist method constrain cluster requir user provid number cluster unknown advanc crucial impact cluster result paper argu natur gener action cluster applic specif constraint decid number cluster purpos introduc cluster model constraint driven cluster cdc find priori unspecifi number compact cluster satisfi user provid constraint gener type constraint consid minimum signific constraint minimum varianc constraint combin type prove np hard cdc problem constraint propos dynam data structur cd tree organ data point leaf node leaf node approxim satisfi cdc constraint minim object function base cd tree develop effici algorithm solv cluster problem experiment evalu synthet real dataset demonstr qualiti gener cluster scalabl algorithm joint cluster analysi attribut relationship data withouta dpriori specif number cluster recognit multi dsentenc dari subcellular local mention biomed abstract method multi drelat classif singl multi dfeatur aggreg function effici process subspac skylin queri high dimension data recent studi effici answer subspac skylin queri separ approach focus pre materi set skylin point subspac focu dynam answer queri set anchor prune skylin point spatial reason effort compress pre materi subspac skylin remov redund storag space approach remain exponenti number dimens queri time approach hand grow substanti data higher dimension prune power anchor weaker paper propos method answer subspac skylin queri high dimension data premateri storag queri time moder propos notion maxim partial domin space maxim partial domin space maxim equal space pair skylin object full space concept foundat answer subspac skylin queri high dimension data queri process involv simpl prune oper skylin comput small subset candid skylin point subspac develop random sampl method comput subspac skylin line fashion extens experi conduct demonstr effici effect method effici mine region outlier spatial data domin neighborhood profit recent research skylin queri attract interest databas data mine commun databas object belong skylin domin respect attribut databas object current method consid call min max attribut price qualiti user minim maxim object spatial attribut coordin repres relev constraint queri result paper introduc skylin queri type take account min max attribut spatial attribut relationship attribut type queri support micro econom approach decis make qualiti cost solut investig altern approach effici queri process symmetr base shelf index structur asymmetr base index structur special purpos extens experiment evalu real dataset synthet dataset demonstr queri type meaning propos algorithm effici scalabl associ rule relat calculu infer nearest rectangl learn revisit abstract nearest rectangl nr learn train instanc gener hyperrectangl queri classifi class nearest rectangl method receiv attent introduct hybrid learner gain accuraci advantag sacrif classif time compar interpret eager learner decis tree paper seek accuraci improv nr learn control gener rectangl infer rectangl infer compact conserv good make local decis experi benchmark dataset valid effect propos approach robust project cluster algorithm discov signific opsm subspac cluster massiv gene express data order preserv submatrix opsm accept biolog meaning subspac cluster model captur gener tendenc gene express subset condit opsm express level gene induc linear order condit opsm mine reduc special case sequenti pattern mine problem pattern support sequenc uniqu opsm cluster small twig cluster long pattern natur low support incur explos comput cost complet prune exist method massiv dataset thousand condit hundr thousand gene common todai gene express analysi interest biologist reveal small group gene tightli coregul condit pathwai process requir gene act concert paper introduc kiwi mine framework massiv dataset exploit paramet provid bias test bound number candid substanti reduc search space problem scale target highli promis seed lead signific cluster twig cluster extens biolog comput evalu real dataset demonstr kiwi effect mine biolog meaning opsm subspac cluster good effici scalabl cluster descript format problem algorithm ultim goal data mine extract knowledg massiv data knowledg ideal repres human comprehens pattern end user gain intuit insight data mine method produc readili understand knowledg cluster algorithm output set point cluster paper perform systemat studi cluster descript gener interpret pattern cluster introduc analyz descript format lead express power motiv defin descript problem trade off interpret accuraci present effect heurist algorithm empir evalu densiti dbase cluster evolv data stream nois cluster import task mine evolv data stream limit memori pass constraint natur evolv data stream impli requir stream cluster assumpt number cluster discoveri cluster arbitrari shape abil handl outlier lot cluster algorithm data stream propos offer solut combin requir paper present denstream approach discov cluster evolv data stream dens micro cluster name core micro cluster introduc summar cluster arbitrari shape potenti core micro cluster outlier micro cluster structur propos maintain distinguish potenti cluster outlier prune strategi design base concept guarante precis weight micro cluster limit memori perform studi number real synthet data set demonstr effect effici method joint cluster analysi attribut data relationship data connect dcenter problem disc dbase approach data summar privaci preserv multidimension subspac skylin analysi skylin oper import multicriteria decis make applic recent studi develop effici method comput skylin object space consid skylin multipl subspac simultan importantli fundament effici process rank queri sweep select abstract exist method top rank queri emploi techniqu includ sort updat threshold materi view paper propos index base techniqu top rank queri index layer skylin index microclust object grid structur develop effici algorithm rank queri locat answer point sweep line hyperplan score function index object method easili plug typic multi dimension databas index comprehens experi demonstr method outperform exist illustr applic data mine techniqu microclust effect solut databas queri process catch view skylin semant approach base decis subspac skylin oper import multicriteria decis make applic recent studi develop effici method comput skylin object specif space fundament problem semant skylin remain open subspac object skylin practic user interest skylin subspac relationship skylin subspac super space effect analyz subspac skylin effici comput skylin subspac paper investig semant skylin propos subspac skylin analysi extend full space skylin comput subspac skylin comput introduc notion skylin group essenti group object coincident skylin subspac identifi decis subspac qualifi skylin group subspac skylin notion concis captur semant structur skylin subspac multidimension roll drilldown analysi introduc develop microeconom data mine problem custom dorient catalog segment mine thick skylin larg databas abstract peopl recent interest oper call skylin return object domin object regard measur multi dimension space recent work skylin oper focus effici comput skylin larg databas work user thin skylin singl object desir real applic paper propos concept call thick skylin recommend skylin object nearbi neighbor distanc effici comput method develop includ effici algorithm sampl prune index estim find thick skylin statist index larg databas highli effici microclust base algorithm mine thick skylin microclusterbas method lead substanti save comput concis represent thick skylin case high cardin experiment perform studi show propos method effici effect top ddown method mine dspecif frequent pattern biolog sequenc accur effici crawl relev websit focus web crawler recent emerg altern establish web search engin focus crawler retriev relev webpag applic target websit singl webpag compani repres websit individu webpag answer queri target websit web directori establish solut paper introduc focus websit crawler emploi paradigm focus crawl search relev websit propos crawler base level architectur crawl strategi explicit concept websit extern crawler view web graph link websit select websit examin invok intern crawler intern crawler view webpag singl websit perform focus page crawl websit experiment evalu demonstr propos focus websit crawler outperform previou method focus crawl adapt retriev websit singl webpag frequent dsubsequ dbase predict outer membran protein number medic import diseas caus bacteria collect call gram neg bacteria note extra outer membran surround cell protein resid membran outer membran protein omp primari research interest antibiot vaccin drug design surfac bacteria access target develop drug develop genom sequenc technolog bioinformat biologist deduc protein produc bacteria attempt classifi protein locat bacteri cell protein local program accur predict omp current develop omp classifi data mine research suggest frequent pattern good perform aid develop accur effici classif algorithm paper present method identifi omp base frequent subsequ test gramneg bacteri protein local determin biolog experi classifi associ rule approach base support vector machin svm compar propos method state art method biolog domain result demonstr method term accur identifi omp provid biolog insight increas understand structur function import protein prefer mine approach mine user prefer person applic abstract advanc person applic requir comprehens knowledg user like dislik order provid individu product recommend person custom advic custom tailor product offer approach model prefer strict partial order semant proven suitabl applic paper present prefer mine techniqu detect strict partial order prefer user log data main advantag approach semant express prefer mine result experiment evalu prove effect effici algorithm prefer mine implement sophist sql statement execut data intens oper databas layer algorithm scale larg log data set approach person applic gain valuabl knowledg custom prefer essenti qualifi custom servic hierarch document cluster frequent itemset major challeng document cluster extrem high dimension vocabulari document set easili thousand word hand document small fraction word vocabulari featur requir special handl requir hierarch cluster cluster document brows increas specif topic paper propos notion frequent itemset associ rule mine document cluster intuit cluster criterion cluster identifi common word call frequent itemset document cluster frequent itemset produc hierarch topic tree cluster focus frequent item dimension document set drastic reduc show method outperform exist method term cluster accuraci scalabl featur weight instanc select collabor filter inform dtheoret approach psort db improv protein subcellular local predict gram dneg bacteria instanc select techniqu memori dbase collabor filter abstract collabor filter cf import data mine techniqu make person recommend book web page movi popular algorithm memori base collabor filter predict user prefer base similar user instanc databas tremend growth user larg number product memori base cf algorithm result problem decid instanc predict order reduc execut cost excess storag possibl improv gener accuraci avoid nois overfit paper focu work typic user prefer databas miss valu propos instanc reduct techniqu call turf turf preprocess step improv effici accuraci memori base cf algorithm kei idea gener predict carefulli select set relev instanc evalu techniqu eachmovi data set experi show propos algorithm dramat speed predict improv accuraci web site mine spot competitor automat extract inform world wide web establish method focu spot singl htmldocument problem spot complet web site handl adequ spite import applic paper discuss classif complet web site point main differ page classif discuss intuit approach weak approach treat web site larg html document appli method page classif show accuraci improv emploi preprocess step assign occur web page topic determin topic repres inform web site classifi accur accomplish direct appli establish classif algorithm featur space occur topic direct treat site tree occur topic markov tree model classif improv effici approach addition introduc power prune method reduc number consid web page experi show superior markov tree approach classif accuraci demonstr prune method reduc process time improv classif accuraci gener term algorithm perform frequent term dbase text cluster text cluster method structur larg set text hypertext document method text cluster address special problem text cluster high dimension data larg size databas understand cluster descript paper introduc approach frequent item term set text cluster frequent set effici discov algorithm associ rule mine cluster base frequent term set measur mutual overlap frequent set respect set support document present algorithm frequent term base text cluster ftc creat flat cluster hftc hierarch cluster experiment evalu classic text document web document demonstr propos algorithm obtain cluster compar qualiti significantli effici state theart text cluster algorithm method provid understand descript discov cluster frequent term set instanc select techniqu memori dbase collabor filter abstract collabor filter cf import data mine techniqu make person recommend book web page movi popular algorithm memori base collabor filter predict user prefer base similar user instanc databas tremend growth user larg number product memori base cf algorithm result problem decid instanc predict order reduc execut cost excess storag possibl improv gener accuraci avoid nois overfit paper focu work typic user prefer databas miss valu propos instanc reduct techniqu call turf turf preprocess step improv effici accuraci memori base cf algorithm kei idea gener predict carefulli select set relev instanc evalu techniqu eachmovi data set experi show propos algorithm dramat speed predict improv accuraci select relev instanc effici accur collabor filter collabor filter databas consum prefer make person product recommend achiev widespread success commerc inform filter applic nowadai tradit collabor filter algorithm scale grow number consum qualiti recommend improv order gain trust consum paper present method improv scalabl accuraci collabor filter algorithm introduc inform theoret approach measur relev consum instanc predict prefer product target concept propos method reduc train data set select highli relev instanc experiment evalu eachmovi data set show method doesn significantli speed predict result accuraci featur weight method improv accuraci collabor filter invit address multipl similar queri basic dbm oper mine metric databas 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 dc dtree fulli dynam index structur data warehous compani recogn strateg import knowledg hidden larg databas built data warehous typic updat collect appli data warehous period batch mode night deriv inform index structur updat standard approach bulk increment updat data warehous drawback averag runtim singl updat small total runtim batch updat larg content data warehous date paper introduc dc tree fulli dynam index structur data warehous model data cube index structur design applic drawback bulk updat approach critic dc tree hierarch index structur similar tree exploit concept hierarchi typic defin effect cooper user comput classif decis tree successfulli task classif state art algorithm incorpor user tree construct process paper present user center approach process user comput contribut strength user domain knowledg evalu intermedi result algorithm comput automat creat pattern satisfi user constraint gener visual pattern cooper approach domain knowledg user direct search algorithm addition provid adequ data knowledg visual pattern recognit capabl human increas effect decis tree construct user deeper understand decis tree obtain result algorithm achiev intend level cooper introduc visual data categor numer attribut techniqu visual decis tree present deep insight process decis tree construct kei contribut integr state art algorithm decis tree construct style cooper rang complet manual combin complet automat classif support experiment perform evalu demonstr cooper approach yield effici construct decis tree small size high accuraci 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 visual classif interact approach decis tree construct satisfi basic requir accuraci understand classifi decis tree classifi popular construct decis tree sophist algorithm introduc fulli interact method base multidimension visual techniqu interact capabl domain knowledg expert profit includ tree construct phase interact construct decis tree user deeper understand data know decis tree gener arbitrari algorithm interact approach overcom limit 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 abstract percept dbase classif decis tree successfulli task classif state art algorithm incorpor user tree construct process paper present user center approach process user comput contribut strength user domain knowledg evalu intermedi result algorithm comput automat creat pattern satisfi user constraint gener visual pattern cooper approach domain knowledg user direct search algorithm addition provid adequ data knowledg visual pattern recognit capabl human increas effect decis tree construct user deeper understand decis tree obtain result algorithm achiev intend level cooper introduc visual data categor numer attribut techniqu visual decis tree present deep insight process decis tree construct kei contribut integr state art algorithm decis tree construct style cooper rang complet manual combin complet automat classif support experiment perform evalu demonstr cooper approach yield effici construct decis tree small size high accuraci increment gener mine data wareh environ data warehous manual analys support visual tool semi automat data mine perform cluster classif summar attribut orient gener common method task summar typic data warehous updat oper collect appli data warehous period deriv inform updat due larg size base relat highli desir perform updat increment paper present algorithm increment attribut orient gener conflict goal good effici minim overli gener algorithm increment insert delet base materi relat intermedi gener level anchor relat experi demonstr increment gener perform effici low degre increment gener mine data wareh environ data warehous manual analys support visual tool semi automat data mine perform cluster classif summar attribut orient gener common method task summar typic data warehous updat oper collect appli data warehous period deriv inform updat due larg size base relat highli desir perform updat increment paper present algorithm increment attribut orient gener conflict goal good effici minim overli gener algorithm increment insert delet base materi relat intermedi gener level anchor relat experi demonstr increment gener perform effici low degre 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 featur base classif protein dock site algorithm larg databas experiment result report develop effici classif algorithm larg databas algorithm classifi protein dock site base set geometr physico chemic surfac featur experi discov classif rule provid interest insight natur dock site speed queri process protein dock databas introduct biochem level process life base protein protein interact protein dock problem attract interest bioinformat commun goal project biowepro develop protein dock system queri protein dock candid extract larg databas protein model protein purpos dock represent properti surfac biowepro queri process di 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 formbasiert such nach komplement ren doberfl chen einer protein ddatenbank databas interfac cluster larg spatial databas knowledg discoveri larg spatial databas focus techniqu effici class identif number size spatial databas rapidli grow larg amount data obtain satellit imag rai crystallographi scientif equip autom knowledg discoveri import spatial databas method knowledg discoveri databas kdd base relat databas system paper address task class identif spatial databas cluster techniqu put special emphasi integr discoveri method db interfac crucial effici kdd larg databas kei integr spatial access method tree focus compon kdd system determin part databas relev knowledg discoveri task present strategi focus select repres spatial databas focus kofi ein expertensystem zur integrierten dokumenten und wissensverwaltung integr knowledg base offic inform system model extend sdm implement prolog 