关键词: 聚类, k-prototypes算法, 混合属性数据, 相异度

Abstract: Based on the K-prototypes, the clustering problem for data mixed with numeric and categorical values was researched in this paper. At first, an improved formula for computing the dissimilarity degree was proposed, compared with the formula in the K-prototypes algorithm. The modified formula can reflect the samples similarities and differences more precisely. Furthermore, a new clustering algorithm for data mixed with numeric and categorical values was presented on the basis of the improved formula for computing the dissimilarity degree, which was finally applied in the clustering analysis of English loanwords. The experimental results show that the new algorithm has better stability and higher precision than the traditional K-prototypes algorithm.

Key words: clustering, k-prototypes algorithm, data with mixed numeric and categorical values, dissimilarity