设为首页 联系我们 加入收藏

当前位置: 网站首页 期刊分类目录 2011第2期 正文

一种改进的模糊C均值聚类算法

作者:时间:2014-11-21点击数:

PDF全文下载:2011020194

刘喜梅, 雷达

 (青岛科技大学 自动化与电子工程学院,山东 青岛266042)

摘要: 针对传统的模糊C均值聚类(FCM)算法聚类数目难以确定,目标函数收敛速度慢的特点,提出了一种改进的模糊聚类算法,将粒度思想和m-α关系引入FCM模糊聚类算法中,从不同的粒度空间对聚类进行有效性评价,并通过改变m或α的值来影响模糊化程度,进而改变聚类的收敛速度。分别采用FCM与该算法对经典数据集进行聚类对比。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和初始聚类中心,并且具有比传统FCM更快的收敛速度。

 关键词: 模糊C均值; 粒度思想; 密度函数; 模糊因子; 收敛速度

 中图分类号: TP 181                   文献标志码: A

 An Improved  Fuzzy C-Means Clustering Algorithm

 LIU Xi-mei,  LEI Da

 (College of Automation and Electronic Engineering, Qingdao University of Science and Technology, Qingdao 266042, China)

 Abstract: Because the Fuzzy C-Means (FCM) clustering algorithm is difficult to determine clustering numbers and has a low convergent speed. The improved FCM fuzzy clustering algorithm is proposed by introducing the granularity thinking and the relationship of m-α into the FCM. The clusters are evaluated from different granular spaces. And the clustering convergent speed is enhanced by changing the values of fuzzy factors m and α to affect the fuzzification degree. The improved FCM algorithm and the FCM are used in classical data sets to make a comparison. The results have shown that the proposed algorithm can obtain reasonable and effective clustering numbers, and has a faster convergent rate than the FCM.

Key words:fuzzy C-means(FCM); granularity thinking; density function; fuzzy factor; convergent speed

收稿日期:2010-11-16

作者简介: 刘喜梅(1961—),女,教授.

Copyright © 2011-2017 青岛科技大学学报 (自然科学版)