A tabu search based algorithm for clustering categorical data sets

Abstract

Clustering methods partition a set of objects into clusters such that objects in the same cluster are more similar to each other than objects in different clusters according to some defined criteria. In this paper, we present an algorithm, called tabu search fuzzy k-modes, to extend the fuzzy k-means paradigm to categorical domains. Using the tabu search based technique, our algorithm can explore the solution space beyond local optimality in order to aim at finding a global optimal solution of the fuzzy clustering problem. It is found that our algorithm performs better, in terms of accuracy, than the fuzzy k-modes algorithm.