Asymmetric Clustering Based on Self-Similarity

Abstract

This paper proposes a clustering method for asymmetric similarity data. In this method, systematic asymmetry in the data is explained by using self-similarity of objects. We exploit an additive fuzzy clustering model for capturing the classification structure in the data. Moreover, the symmetric similarity data is restored by using the result of the clustering method. Therefore, we can exploit many data analyses in which objective data is symmetric similarity data. Several numerical examples are shown in order to show the better performance of the proposed method.