An Extended Fuzzy k-Means Algorithm for Clustering Categorical Valued Data
Wang Jiacai, Gu Ruijun · 2010
Although fuzzy k-modes algorithm has removed the numeric-only limitation of the k-means algorithm, that each attribute of the centroid with a single category value and the use of a simple distance measure will compromise its precision, and therefore prone to falling into local optima. In this paper, an extended fuzzy k-means(xFKM) algorithm for clustering categorical valued data is presented, in which the cluster centroid vectors are represented as expanded forms to keep more clustering information as possible as well, and updated with the method similar to fuzzy k-means algorithm. Experiments on several real databases show that xFKM algorithm can get better clustering result than fuzzy k-modes algorithm does.