一种改进的k-modes聚类算法

传统的K-modes算法采用了简单的0-1匹配来计算属性间的相异度,后改进为频率计算相异度,但是他们都忽略了各属性间的差异。本文研究了基于粗糙集和知识粒度的属性加权算法,该算法既克服了属性的冗余问题又综合考虑了各属性间的差异。在此基础上,通过对传统K-modes算法进行属性加权来改进K-modes算法中忽略的属性间差异问题。通过与其他的K-Modes算法进行实验比较,结果表明新的算法更加有效的。

An Improved K-Modes Clustering Algorithm

The traditional K-modes algorithm, the simplematching dissimilarity measure, is used to compute the distance between two values of the samecategorical at tributes. This compares two categorical values directly and results in either a differenceof zero when the two values are identical or one if otherwise. However it ignores the differences among the attributes. In this paper, we studyan attribute weighting algorithm based on rough set and knowledge granulation. This algorithm not only overcomes the redundancy of attributes, but also takes into account the differences among attributes. Attributes weightingin the traditional K-modes algorithm are used to improve the K-modes algorithm to ignore the difference between attributes. Compared with other K-Modes clustering algorithms, the results show that the new algorithm is more effective.