计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (22): 172-174.DOI: 10.3778/j.issn.1002-8331.2008.22.051

• 数据库、信号与信息处理 • 上一篇    下一篇

软硬结合的快速模糊C-均值聚类算法的研究

尹海丽1,王颖洁2,白凤波3   

  1. 1.青岛理工大学,山东 青岛 266033
    2.大连大学 信息工程学院,辽宁 大连 116622
    3.海辉软件(大连)有限公司,辽宁 大连 116023
  • 收稿日期:2008-05-06 修回日期:2008-06-24 出版日期:2008-07-11 发布日期:2008-07-11
  • 通讯作者: 尹海丽

Research of fast fuzzy C-means clustering algorithm based on soft and hard clustering

YIN Hai-li1,WANG Ying-jie2,BAI Feng-bo3   

  1. 1.Qingdao University of Technology,Qingdao,Shandong 266033,China
    2.Information Science and Engineering Institute of Dalian University,Dalian,Liaoning 116622,China
    3.hiSoft Technology (Dalian) Co.,Ltd,Dalian,Liaoning 116023,China
  • Received:2008-05-06 Revised:2008-06-24 Online:2008-07-11 Published:2008-07-11
  • Contact: YIN Hai-li

摘要: 讨论的是对模糊C-均值聚类方法的改进,在原有的模糊C-均值算法的基础上,提出一种软硬结合的快速模糊C-均值聚类算法。快速模糊C-均值聚类算法是在模糊C-均值聚类算法之前加入一层硬C-均值聚类算法。硬聚类算法能比模糊聚类算法以高得多的速度完成,将硬聚类中心作为模糊聚类中心的迭代初值,从而提高模糊C-均值聚类算法的收敛速度,这对于大量数据的聚类是很有意义的。用数据仿真验证了这种快速模糊C-均值聚类算法比模糊C-均值算法迭代调整过程短,收敛速度快,聚类效果好。

关键词: 模糊C-均值算法, 模糊聚类, 软聚类, 硬聚类

Abstract: This paper discusses how to improve the fuzzy C-means clustering algorithm(FCM).On the basis of FCM,the paper puts forward a kind of fast fuzzy C-means clustering algorithm based on soft and hard clustering.The fast FCM inserts one layer of hard C-means clustering algorithm in front of FCM.The hard C-means clustering algorithm can be finished at much higher speed than FCM.In order to improve the convergence speed of FCM,the authors regard the cluster centers of hard C-means clustering algorithm as the initial value of fuzzy cluster centers.It is very meaningful for a large number of data clustering.In addition,this paper proves that the fast fuzzy C-means clustering algorithm has a shorter adjusting iteration course and a faster convergence speed than FCM and the clustering result achieved from data emulation is very ideal.

Key words: fuzzy C-means clustering algorithm(FCM), fuzzy clustering, soft clustering, hard clustering