Abstract: Traditional fuzzy K-Modes clustering algorithm uses a simple matching dissimilarity measure to compute the dissimilarity between an object and Mode. However, how well Mode is representative of the cluster is not considered in the dissimilarity measure, which may lose some information and result in the cluster with weak intra-similarity. This paper proposes a new dissimilarity measure between an object and Mode, which uses membership degrees of objects to clusters to reflect how well Mode is representative of the cluster. Comparisons with traditional K-Modes and fuzzy K-Modes algorithm illustrate the effectiveness of the new distance measure.

Key words: fuzzy K-modes clustering algorithm, dissimilarity measure, cluster center

摘要： 传统的模糊K-Modes聚类算法采用简单匹配方法度量对象与Mode之间的相异程度，没有充分考虑Mode对类的代表程度，容易造成信息的丢失，弱化了类内的相似性。针对上述问题，通过对象对类的隶属度反映Mode对类的代表程度，提出一种新的相异度量，并将它应用于传统的模糊K-Modes聚类算法。与传统的K-Modes和模糊K-Modes聚类算法相比，该相异度量是有效的。

关键词: 模糊K-Modes聚类算法, 相异度量, 类中心

CLC Number: 

• TP311