离差最大化特征加权模糊c-划分的聚类分析

Clustering Analysis of Maximizing Deviation Feature Weighting Fuzzy C-partitions

聚类分析中每个样本用一个m维向量来表征,每个分量描述样本的一个特征,由于对特征的提取不够完善,使得m维向量的每个分量对聚类贡献不均。基于此本文利用向量间离差最大化对样本的每个分量即特征进行加权,提出一种新的加权模糊c-划分的聚类分析法,一定程度上克服了模糊c-划分的聚类分析对每个特征等同对待不足,又保持其算法的收敛性,最后给出一个算例说明此算法的优越性。

Every sample of clusteringanalysis can be represented by m dimensional vector, and every component of vector describes a feature of sample. Because the extraction of feature is imperfect, so the attribution of every component in m dimensional vector is totally different. Based on that, this paper utilize the maximizing deviation of vector to weight every component, ie, feature of sample, and a new fuzzy c-partitions clustering analysis has been proposed. To some extent, the method is not only overcome limation which treats every feature equally in fuzzy c-partitions clustering analysis, but also keeps the convergence of algorithm. Finally, an example is also given to show the superiority of algorithm.