计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (32): 143-147.DOI: 10.3778/j.issn.1002-8331.2008.32.043

• 数据库、信号与信息处理 • 上一篇    下一篇

基于网格生长树的微粒群聚类算法

曹明华,曾建潮,介 婧   

  1. 太原科技大学 系统仿真与计算机应用研究所,太原 030024
  • 收稿日期:2007-12-11 修回日期:2008-02-29 出版日期:2008-11-11 发布日期:2008-11-11
  • 通讯作者: 曹明华

Research of PSO-clustering algorithm based on Gird-Propagating Tree

CAO Ming-hua,ZENG Jian-chao,JIE Jing   

  1. Division of System Simulation and Computer Application,Taiyuan University of Science and Technology,Taiyuan 030024,China
  • Received:2007-12-11 Revised:2008-02-29 Online:2008-11-11 Published:2008-11-11
  • Contact: CAO Ming-hua

摘要: 提出了一种基于网格生长树的微粒群聚类算法。算法利用网格和密度阈值去除数据集中的孤立点,从网格集中随机地选取种子点,以基于密度距离作为判断生长方向及分类的依据,以网格生长树的大小作为聚类目标函数。引入微粒群算法确定最终的聚类结果。测试表明,基于网格生长树的微粒群聚类算法对于大规模形状复杂非重叠的数据是可行且有效的。

关键词: 聚类算法, 网格生长树, 微粒群算法

Abstract: A PSO-clustering algorithm based on Gird-Propagating Tree(PSO-GPT) is presented in this paper.This algorithm uses the gird and density threshold method to remove isolated points of data set,selects the initial seed points in propagating tree randomly,judges vegetal direction and executes sort according to the density-based distance,calculates cluster aim function by the propagating tree value.PSO algorithm is employed to determine the final clustering results in the new algorithm.Experimental results show that the proposed method is feasible and effective for large-scale complex shape and non-duplication data.

Key words: clustering algorithm, Gird-Propagating Tree, Particle Swarm Optimization(PSO)