一种有效的k_means聚类初始中心选取方法.pdfVIP

2010 7 JISUANJI YU XIANDA IHUA 179 : 2010) kmeans , (天津工业大学, 天津 300160) :针对传统 km eans聚类算法中对初始聚类中心随意选取和人为指定的缺陷, 提 一种改进的初始聚类中心的选取 方法, 利用差异矩阵将新的聚类初始中心计算方法用在传统的 km eans算法思想中, 对传统的 km eans算法进行改进 降低 km eans算法的复杂度和对异常点的敏感度, 提高算法的可伸缩性 : kmeans; 聚类; 初始化中心; 差异矩阵 : P301. 6 : A doi: 10. 3969/j. issn. 2010. 07. 023 A n E ffectiv eM ethod for In itial Cen trepoin ts of K m ean s Clu stering REN Jingbiao, Y IN Shaohong ( ianjin Polytechnic University, ianjin 300160, China) A b stract: In this paper, a mi proved in itial cluster center selection m ethod is presented for the defects of the traditional kmeans cluster algorithm on the initial cluster center selection, and uses the new cluster initial center calcu lation m ethod on the m ind of trad itional km eans algorithm. It is show n that the new algorithm not only reduces the comp lexity of kmeans algorithm, but also mi proves the scalab ility of the km eans algorithm. K ey w ords: km eans algorithm; clustering; in itial centerpoints; d issmi ilarity m atrix : (KDD), 0 , ( clustering) [ 1] 1 kmeans : X, C X kmeans, , CX C, C , C , , C : , 1 2 K ( 1) C C C = X; O ( nkt), k, t 1 2 k ( 2) C C = ( i j) i j , n, km eans , , kmeans , , , ; , , kmeans, , , [ 7] , , k