Abstract: A clustering method based on genetic algorithm is presented. The cluster centers are binary encoded. The sum of the Euclidean distances of the points from their respective cluster centers is adopted as the similarity metric. The optimal cluster centers are searched by selection, crossover and mutation. Experimental results demonstrate that GA-based clustering is better than K-means algorithm.

摘要： 给出了一种基于遗传算法的聚类分析方法。采用二进制编码方式对聚类的中心进行编码，并用特征向量与相应聚类中心的欧氏距离的和来判断聚类划分的质量，通过选择、交叉和变异操作对聚类中心的编码进行优化，得到使聚类划分效果最好的聚类中心。实验结果显示，该方法的聚类划分效果明显优于传统的K-均值方法。

关键词: 遗传算法, 聚类, K-均值算法, 二进制编码

CLC Number: 

• TP18