Rough K-modes Clustering Algorithm Based on Entropy

Cluster analysis is an important technique used in data mining. Categorical data clustering has received a great deal of attention in recent years. Some existing algorithms for clustering categorical data do not consider the importance of attributes for clustering, thereby reducing the efficiency of clustering analysis and limiting its application. In this paper, we propose a novel rough k-modes clustering algorithm based on entropy. First, we integrated the knowledge of information entropy to define a new dissimilarity measure that takes into account the importance of attributes for clustering and improves the quality of clustering. Then, applying the theory of rough set analysis, we used upper and lower approximation to deal with uncertain clusters, which allowed us to offer an improved solution for uncertainty analysis. Finally, our experimental results demonstrated that our proposed algorithm performed better than other conventional clustering algorithms in terms of clustering accuracy, purity, and F1-measure.