Abstract: A mountain clustering algorithm based on particle swarm optimization is constructed in this paper by combining the two algorithm of mountain clustering and particle swarm optimization. Firstly, the grid is built in the data space, then a mountain function denoting the density of data is made, and lastly, the operation of razing out the mountains orderly to find clustering centers is replaced by NichePSO algorithm. All global maximums of mountain function are found by running NichPSO algorithm, and the centers number and position are found at same time. The simulating experiments indicate that the new algorithm can offset some limitation of traditional clustering algorithms.

Key words: Particle swarm optimization, Clustering, Mountain function, Niching

摘要： 将山峰聚类法和小生境微粒群算法结合,构建一种基于小生境微粒群算法的山峰聚类法: 首先在数据空间上构造网格, 进而构造出表示数据密度指标的山峰函数, 然后将山峰聚类方法中通过顺序地削去山峰函数来选择聚类中心这一步用小生境微粒群算法代替, 通过执行小生境微粒群算法对山峰函数进行多峰函数寻优, 找到山峰函数的每一个峰, 即可确定聚类中心的个数和每一个聚类中心位置. 仿真实验表明, 构建的新算法能够弥补传统聚类算法的一些缺陷.

关键词: 微粒群算法, 聚类, 山峰函数, 小生境