J4

• 计算机科学 • 上一篇    下一篇

一种基于核的模糊聚类算法

曲福恒1, 马驷良1, 胡雅婷2   

  1. 1. 吉林大学 数学研究所, 长春 130012; 2. 吉林农业大学 信息技术学院, 长春 130118
  • 收稿日期:2008-02-04 修回日期:1900-01-01 出版日期:2008-11-26 发布日期:2008-11-26
  • 通讯作者: 马驷良

A Kernel Based Fuzzy Clustering Algorithm

QU Fuheng1, MA Siliang1, HU Yating2   

  1. 1. Institute of Mathematics, Jilin University, Changchun 130012, China;2. College of Information and Technology, Jilin Agriculture University, Changchun 130118, China
  • Received:2008-02-04 Revised:1900-01-01 Online:2008-11-26 Published:2008-11-26
  • Contact: MA Siliang

摘要: 结合核技术与改进的模糊c均值算法聚类准则提出一 种基于核的模糊聚类算法. 通过引入核函数, 样本点被非线性变换映射到高维特征空间进行聚类, 提高了聚类性能. 同时, 算法改进了模糊c-均值聚类模型中的概率型约束条件, 使其对噪声和野值点具有较好的鲁棒性. 在真实数据和人造数据上与常用聚类算法进行了对比实验, 结果表明该算法具有较低的时间、 空间复杂度与较好的聚类性能.

关键词: 聚类分析, 核函数, 模糊c-均值, 特征空间

Abstract: A new kernel based fuzzy clustering algorithm was proposed on the basis of combining the kernel technique with the rules of the im proved fuzzy c-means clustering algorithm. In the proposed algorithm, the sample points are mapped into the feature space via introducing the kernel function into the clustering model to improve the performance. The new algorithm is robust to the noises because it relaxes the constraint conditions used in the fuzzy c-means clustering model. We compared the results with those of some most frequently used clustering algorithms to show the effectiveness of the proposed algorithm.

Key words: cluster analysis, kernel function, fuzzy c-means, feature space

中图分类号: 

  • TP391.4