邻域平衡密度聚类算法.pdfVIP

计算机研究与发展 ISSN N 111777TP J ournal of Comp ut er Research and Develop ment 47 (6) : 1044 1052 , 20 10 邻域平衡密度聚类算法 武佳薇 　李雄飞 　孙 　涛 　李 　巍 ( ( ) ) 符号计算与知识工程教育部重点实验室 吉林大学 　长春 　1300 12 (wj w725 1 @163 . com) A DensityBased Cl ustering Algorithm Concerning Neighborhood Balance Wu J iawei , Li Xiongf ei , Sun Tao , and Li Wei ( Key L aboratory of S y m bol Comp ut ation an d K now le d g e Eng i nee ri ng ( J i l i n Uni ve rs i ty ) , M i nis t ryof E d ucation , Chang ch un 1300 12) Abstract 　Clu st erin g i s an import ant analytical tool in dat a minin g . Den sit yba sed clu st ering analy si s i s a clu st ering analy si s met ho d w hich i s demanded to deal wit h very lar ge dat aba se s. By analyzin g t he limit ation of t he exi stin g den sit yba sed clu st erin g algorit hm s an d t he p roblem s of di spo sin g variou s den sitie s of dat a and illegibilit y of clu st er s boundarie s , definition s such a s p roj ectio n point s , neighbor hoo d balance , balanceable core p oint s , and boundary sp ar se point s are int ro duced . Aft er analyzing t he di st ribution charact er s of core point s and p oint s in t heir neighbor hoo d , a den sit y ba sed clu st erin g algorit hm bDB SCAN concer ning t he neighbor hoo d balance of core point s i s p ropo sed to imp rove DB SCAN . The algorit hm deal s wit h t he core p oint s by get ting t he p roj ection of t he point s in t heir neighbor hoo d to j udge w het her t hey are balanceable . Only balanceable core point s can be exp an ded to for m clu st er s. The algorit hm can di scover clu st er s wit