初始聚类中心优化选取的核C-均值聚类算法

在通常的核C-均值聚类算法中,聚类结果对初始聚类中心敏感,随机选取初始聚类中心时,会使得迭代次数较多、分类结果不稳定.针对该问题提出一种优化选取仞始聚类中心的算法.该方法采用一种改进的最大最小距离算法对原始空间中的数据进行粗分类,将粗分类结果中每类类心作为初始聚类中心,再运用核C-均值聚类算法进行分类.仿真实验结果表明方法能有效减少迭代次数,使分类结果更加稳定,分类识别率也有一定程度的提高.

A Kernel C-Means Clustering Algorithm with Meliorated Initial Center

The kernel C-means clustering algorithm is sensitive to the initial start centers.As the initial clustering centers are randomly chosen,the iterative amount is large and the result of the classification is unstable.To solve this problem,a new method is proposed to find the meliorated initial clustering center.Firstly a kind of improved max-min distance means is used to classify the original data,then the centers in the observed space are obtained as the meliorated initial centers.At the end,kernel C-means c...