摘要： 提出一种求解聚类问题的分布估计算法。基于PBIL算法定义聚类矩阵，建立对应的概率矩阵模型，引入遗传算法的基因变异算子，设计适用于分布估计算法的变异操作，改进概率模型的更新方式。实验结果表明，与Kmeans、Kmedioid、Clarans和遗传算法相比，该算法的聚类质量较好。

关键词: 聚类, 分布估计算法, 概率模型, 遗传算法, 变异

Abstract: Based on PBIL algorithm, clustering matrix and clustering probability matrix model are designed. After leading in arithmetic operators of gene mutation in genetic algorithm, mutation which adapts to Estimation of Distribution Algorithms(EDA) is designed. After improving update mode of probabilistic model, a new estimation of distribution algorithms which can be applied to solving clustering problems is originated. By comparing new algorithms with Kmeans, Kmedioid, Clarans algorithm and Genetic Algorithm(GA), better clustering quality and general performance of new algorithms are confirmed by experimental results.

Key words: clustering, Estimation of Distribution Algorithm(EDA), probabilistic model, Genetic Algorithm(GA), mutation

中图分类号: 

• TP18