一种改进的FCM聚类研究  

Analysis of an Improved Fuzzy C-Means Clustering

作　　者：杨璀琼[1] 江红[1] 余晓磊[1] 

机构地区：[1]华东师范大学计算中心,上海200062

出　　处：《计算机与数字工程》2010年第5期1-3,15,共4页Computer & Digital Engineering

摘　　要：针对模糊C-均值(FCM)算法在解决各数据样本对聚类中心具有同样影响权重问题的不足以及对噪声和孤立点数据敏感,提出了改进和提高的方法:利用相似关系理论,为每一个样本加一个特征权值,构造加权目标函数,让不同的样本在聚类中起不同的作用,并对欧式距离进行加权,减少少数异常点对确定聚类中心的影响;同时对隶属度函数进行改进,以消除孤立点对聚类结果的影响。实验结果表明改进后的算法比经典FCM具有更好的鲁棒性和聚类效果。For the problems of the various data sample has the same impact of weight on the clustering center and its sensitivity for data of outlier and noise,the fuzzy C-Means has deficiency.This paper presents the corresponding methods for improving and enhancing： adds a feature weight to each sample utilizing the similar relations theory and constructs the weighting objective function to allow different samples to play different roles in clustering,and weights the Euclidean distance to reduce the impact of small number of abnormal points while determining the cluster centers.Meanwhile,by improving the subject function,the impacts of outliner are eliminated.The experimental results show that the modified algorithm is more robust and has higher clustering accuracy.

关 键 词：模糊C-均值 相似关系 特征权值 欧式距离 隶属度函数 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]