吉林大学学报(理学版)

• 计算机科学 • 上一篇    下一篇

改进的FCM半监督聚类算法

郭新辰 1, 樊秀玲 1, 郗仙田 1, 韩啸2   

  1. 1. 东北电力大学 理学院, 吉林 吉林 132012; 2. 吉林大学 学报编辑部, 长春 130012
  • 收稿日期:2014-01-10 出版日期:2014-11-26 发布日期:2014-12-11
  • 通讯作者: 韩啸 E-mail:hanxiao@jlu.edu.cn

Improved Fuzzy C-Means Clustering Algorithm

GUO Xinchen 1, FAN Xiuling 1, XI Xiantian 1, HAN Xiao2   

  1. 1. College of Science, Northeast Dianli University, Jilin 132012, Jilin Province, China;2. Editorial Department of Journal of Jilin University, Changchun 130012, China
  • Received:2014-01-10 Online:2014-11-26 Published:2014-12-11
  • Contact: hanxiao E-mail:hanxiao@jlu.edu.cn

摘要:

通过将类间分离度函数引入到模糊C-均值聚类算法中, 结合半监督的思想, 建立基于信息熵的半监督模糊C-均值聚类模型, 并对该模型的求解过程进行推导, 提出一种新的算法. 为了验证算法的有效性, 将该算法在UCI数据集上进行实验, 实验结果表明, 该算法比仅引入信息熵的模糊C-均值聚类方法聚类性能更好.

关键词: 半监督聚类, 模糊C-均值算法, 信息熵

Abstract:

A new fuzzy C-means clustering algorithm was proposed by  the introduction of functions of separation between clusters into FCM clustering algorithm and with the nature of semisupervised learning considered. The model of semisupervised FCM clustering algorithm with the information entropy as constraints was established and the solution to the model was derived. The simulation experiments were performed on UCI data sets to verify the effectiveness of the proposed algorithm. The experimental results show that this modified algorithm gets the better validity and performance.

Key words: semisupervised clustering, fuzzy C-means algorithm (FCM), information entropy

中图分类号: 

  • TP181