Adaptive Resonance Theory-based Clustering for Handling Mixed Data

Abstract

This paper proposes an Adaptive Resonance Theory (ART)-based clustering algorithm for a dataset which contains numerical and categorical attributes simultaneously. In the proposed algorithm, similarity between numerical attributes is calculated by the correntropy-based nonlinear similarity measurement, while similarity between categorical attributes is defined by a hamming distance-based approach. One advantage of the proposed algorithm is that the algorithm continually and adaptively generates a sufficient number of nodes for clustering from given data points. Empirical studies on various datasets show that the proposed algorithm has comparable clustering performance to the representative mixed data clustering algorithms.