资源描述：
《基于网格的密度树算法》由会员上传分享，免费在线阅读，更多相关内容在行业资料-天天文库。
1、ProceedingsoftheFourthInternationalConferenceonMachineLearningandCybernetics,Guangzhou,18-21August2005ACLUSTERINGALGORITHMBASEDONBUILDINGADENSITY-TREE1112WEI-DIDAI,YUE-XIANHOU,PI-LIANHE,XIAO-SHENZHENG1DepartmentofComputerScienceandTechnology,TianjinUniversity,Tianjin300072,Chin
2、a2CollegeofMarineScienceandEngineering,TianjinUniversityofScienceandTechnology,Tianjin300457,ChinaE-MAIL:davidy@126.com,yxhou@tju.edu.cn,plhe@tju.edu.cn,zhengxiaoshen@163.comAbstract:formalldatapoints.ComparingwithotherclusteringAnewkindofclusteringalgorithmcalledCABDETisalgori
3、thms,DENCLUEgeneralizespartition-basedmethodspresentedinthispaper.CABDETcreatesatreestructureforandhierarchicalmethods,andhasgoodclusteringeverycluster,fromwhichtheneighbor’sradiusofthecurrentpropertiesfordatasetswithlargeamountsofnoise.objectiscalculatedbythelocaldensityofitsf
4、athernode.However,themethodissensitivetotheinputparameterandThoseunprocessedobjectsintheneighborofthecurrentnoisethreshold.Theselectionofsuchparametersmayobjectareaddedtoextendthetreestructureuntilnonewsignificantlyinfluencethequalityoftheclusteringresults.objectisfounded.Eachd
5、ensity-treeisregardedasonecluster.CABDETrequiresonlyoneinputparameterastheinitialDBSCANcanquicklydiscoverdifferentclustersofradiusoftherootnodeandhasnolimitationofdensityarbitraryshapeaccordingtotheconnectivityofdensity.Athreshold.Othercharacteristicsincludetheabilitiesofcluste
6、risgeneratedorcombinedtoanotherexitingclusterifdiscoveringclusterswitharbitraryshapeandprocessingthethisclusterhasnotlessthanonecoreobjectandmanynoisedata.Theresultofourexperimentsdemonstratesthatdensity-reachableobjects.AnewclusterisfoundbasedonCABDETissignificantlymoreaccurat
7、eindiscoveringsuchfacts:oneclustercanbedeterminedbyanycoreobject.density-changeableclusteringthanthealgorithmDBSCAN,DBSCANrequirestwoinputparameters:radiusoftheandthatCABDETislesssensitivetoinputparameters.ε-neighborhoodandtheminimumnumberofobjects.Actually,high-dimensionalreal
8、datasetsoftenhaveveryKeywords:skeweddistributionssucht
