电子学报 ›› 2012, Vol. 40 ›› Issue (2): 254-259.DOI: 10.3969/j.issn.0372-2112.2012.02.008

• 学术论文 • 上一篇    下一篇

一种基于动态遗传算法的聚类新方法

何宏, 谭永红   

  1. 上海师范大学信息与机电工程学院,上海 200234
  • 收稿日期:2010-10-15 修回日期:2011-08-28 出版日期:2012-02-25 发布日期:2012-02-25

A Novel Clustering Method Based on Dynamic Genetic Algorithm

HE Hong, TAN Yong-hong   

  1. College of Information,Mechanical and Electronic Engineering,Shanghai Normal University,Shanghai 200234,China
  • Received:2010-10-15 Revised:2011-08-28 Online:2012-02-25 Published:2012-02-25

摘要: 如何确定聚类数目一直是聚类分析中的难点问题.为此本文提出了一种基于动态遗传算法的聚类新方法,该方法采用最大属性值范围划分法克服划分聚类算法对初始值的敏感性,并运用两阶段的动态选择和变异策略,使选择概率和变异率跟随种群的聚类数目一致性变化,先进行不同聚类数目的并行搜索,再获取最优的聚类中心.七组数据聚类实验证明该方法能够实现数据集最佳划分的自动全局搜索,同时搜索到最佳聚类数目和最佳聚类中心.

关键词: 聚类分析, 遗传算法, 动态选择, 变异

Abstract: How to determine the number of clusters is always a difficult problem in data cluster analysis.Therefore,a novel dynamic genetic clustering algorithm (DGCA) is proposed in this paper.The DGCA adopts a maximum attribute range partition method to overcome the sensitiveness to initial values of cluster centers for clustering algorithms.Furthermore,the two-stage dynamic selection and mutation operations are used in the DGCA to make selection probability and mutation probability vary with the consistency of the number of clusters in the population.Firstly the parallel search in different numbers of clusters is carried out.Then the optimal search for the best cluster centers is conducted.Numerical experiments on seven data sets show that the proposed DGCA can realize the global search for the best partition and find the optimal values for both the number of clusters and the cluster centers.

Key words: cluster analysis, genetic algorithm, dynamic selection, mutation

中图分类号: