• 中国计算机学会会刊
  • 中国科技核心期刊
  • 中文核心期刊

J4 ›› 2007, Vol. 29 ›› Issue (1): 94-96.

• 论文 • 上一篇    下一篇

利用多群体PSO算法生成分类规则

延丽平[1] 曾建潮[2]   

  • 出版日期:2007-01-01 发布日期:2010-05-30

  • Online:2007-01-01 Published:2010-05-30

摘要:

本文通过对PSO算法模型和分类模型的分析,提出了应用多群体PSO算法实现分类规则的方法。这种方法将c(c≥2)类问题看成是c个两类问题,应用c个微粒群表示c类规则,每 个微粒群应用PSO算法实现对连续变量空间的分类。最后,在五个数据集上的实验结果表明了此方法的可行性和有效性,并与C4.5算法的结果进行了比较。

关键词: PSO算法 分类规则 混合编码

Abstract:

A new approach for classification rules using multi-particle swarm is proposed.In this way,the c(c≥2) classification problem is taken as c two-class problems,and c class rules are expressed by c particle swarms, and classfication rules with the continuous variable space is finished by the PSO algor ithm.Experimental results on 5 data sets show its efficiency and feasibility and are compared with the C4.5 results.

Key words: PSO algorithm;classification rule;hybrid code