计算机应用 ›› 2010, Vol. 30 ›› Issue (05): 1277-1279.

• 数据挖掘与人工智能 • 上一篇    下一篇

一种改进的模糊聚类算法

周红芳1,宋姣姣2,罗作民3   

  1. 1. 西安理工大学
    2.
    3. 西安理工大学计算机科学与工程学院
  • 收稿日期:2009-11-11 修回日期:2010-01-07 发布日期:2010-05-04 出版日期:2010-05-01
  • 通讯作者: 周红芳
  • 基金资助:
    国家863计划项目;陕西省自然科学基础研究计划项目;陕西省教育厅科学研究计划资助项目

Improved fuzzy clustering algorithm

  • Received:2009-11-11 Revised:2010-01-07 Online:2010-05-04 Published:2010-05-01

摘要: 传统模糊聚类算法如模糊C-均值(FCM)算法中,用户必须预先指定聚类类别数C,且目标函数收敛速度过慢。为此,将粒度分析原理应用在FCM算法中,提出了基于粒度原理确定聚类类别数的方法,并采用密度函数法初始化聚类中心。实验结果表明,改进后的聚类算法能够得到合理有效的聚类数目,并且与随机初始化相比,迭代次数明显减少,收敛速度明显加快。

关键词: 模糊C-均值, 粒度分析原理, 耦合度, 分离度, 密度函数

Abstract: In traditional Fuzzy C-Means (FCM) algorithm, the user must give the number of clusters in advance and the objective function converges slowly. To solve these problems, a new algorithm for finding the best number of clusters was proposed with introducing granular analysis principle into FCM clustering algorithm, and density function algorithm was adopted to initialize the cluster centers. The experimental results show that the proposed algorithm can obtain reasonable and effective number of clusters. Compared with the stochastic initialization, this algorithm has fewer iterations and has faster speed to converge.

Key words: Fuzzy C-Means (FCM), granular analysis principle, coupling degree, separating degree, density function