unsupervis class separ multivari data cumul varianc dbase rank resolut dbase outlier factor detect top outli data point engin data estim true fals posit rate higher dimension problem data mine applic estim accuraci observ estim true fals posit rate seri sampl high dimension data mine problem date issu larg neglect previous algorithm provid facilit comput involv high dimension data mine task increas sparsiti lead decreas true posit rate estim effect estim true size membership class cluster allow identifi top candid fals neg track likelihood fals posit estim true fals posit rate research avoid unnecessari cost collect number sampl need propos algorithm comput nonparametr outlier detect effect discov top dn outlier engin data abstract present resolut base outlier notion nonparametr outlier mine algorithm effici identifi top list outlier wide varieti dataset algorithm gener reason outlier result take local global featur dataset consider experi conduct synthet dataset real life construct equip dataset larg build contractor comparison current outlier mine algorithm propos algorithm effect migrat legaci applic palm platform parameterless method effici discov cluster arbitrari shape larg dataset cluster problem group data base similar consist maxim intra group similar minim inter group similar problem cluster data set unsupervis classif class label exist cluster algorithm requir paramet steer cluster process famou number expect cluster constitut supervis sort present paper effici fast scalabl cluster algorithm cluster rang resolut find potenti optimum cluster requir paramet input experi show algorithm outperform exist cluster algorithm qualiti speed larg data set data cluster analysi scalabl constraint valid paper discuss recent cluster approach recount experi algorithm present problem cluster presenc constraint discuss issu cluster valid 