Multi-label Classification Based on Adaptive Resonance Theory

Naoki Masuyama, Yusuke Nojima, Chu Kiong Loo, Hisao Ishibuchi

Abstract: This paper proposes a multi-label classification algorithm based on an algorithm adaptation approach by applying the Adaptive Resonance Theory (ART) and the Bayesian approach for a label association process. In the proposed algorithm, the prior probability and likelihood are updated sequentially. Moreover, an ART-based clustering algorithm continually extracts useful information for multi-label classification, and holds the extracted information on prototype nodes generated by the clustering algorithm. Thanks to the above properties, the proposed algorithm can continually learn multi-label data. Our experimental results in this paper show that the proposed algorithm has better classification performance compared to typical multi-label classification algorithms.