摘要： 聚类是数据挖掘的主要技术之一，是一种无导师监督的模式识别方式。聚类分析是按照数据间的相似程度，依据特定的准则将数据划分成不同子类。该文结合微粒群(PSO)算法，提出一种数字属性聚类算法，为避免PSO算法可能出现的早熟问题，引入混沌的思想，同时考虑到各个聚类的内部相似的特性，将空间特性引入到PSO算法中。仿真实验表明，该算法在解决数字属性聚类的问题上有着良好的性能。

关键词: 群体智能, 聚类算法, 混沌, 空间特性微粒群算法

Abstract: Clustering is one of main technical of data mining, a kind of non-teacher supervises recognition pattern. The clustering analysis concerns about the similar degree of data and rests on the specific criterion to divide the data to the different subclass. This paper unifies the Particle Swarm Optimization(PSO) algorithm, proposes a numeric clustering algorithm. In order to avoid the precocious problem which PSO algorithm possibly appears, it uses the chaos idea, simultaneously, considering each cluster’s internal similar characteristic, also introduces the spatial idea to the PSO algorithm. Experimental result indicates that the simulation algorithm has good performance in the numeric attribute cluster problems.

Key words: swarm intelligence, clustering algorithm, chaos, PSO algorithm with spatial characteristic

中图分类号: 

• TP18