聚类分析中非0、1编码的遗传算法  

Improve K-means Algorithms of Cluster Method By GA with Non 0、1 Code

作　　者：唐立新[1] 杨自厚[1] 王梦光[1] 

机构地区：[1]东北大学系统工程研究所,沈阳110006

出　　处：《应用数学与计算数学学报》1998年第1期22-28,共7页Communication on Applied Mathematics and Computation

摘　　要：K-平均算法属于聚类分析中的动态聚类法,但其聚类效果受初始聚类分类或初始点的影响较大。本文提出一种遗传算法(GA)来进行近代初始分类,以内部聚类准则作为评价指标,实验结果表明,该算法明显好于K-平均算法。K-means algorithms belongs to partitional methods.For clustering.The K-means algorithms for clustering is very much dependent on the initial seed valves. We use a genetic algorithms with non 0, 1 code to make initial near optimal clustering in the k-means algorithm. In our test resulths the GA outperformed the stand k-means algorithms when internal clustering criterion are computed as evalnaction measures. , . ??

关 键 词：聚类分析 K-平均算法 近优初始分类 遗传算法 

分 类 号：O212.4[理学—概率论与数理统计]