Adaptive Resonance Theory-Based Topological Clustering with Node Deletion Mechanism for Evolving Stream Data

Concept drift where the data distribution changes over time deteriorates the performance of machine learning models. Adaptive Resonance Theory (ART)-based clustering algorithms are capable of learning continually and achieving fast and stable clustering. However, concept drift has not been actively discussed in ART-based clustering algorithms. In this paper, we propose an ART-based clustering algorithm that can deal with concept drift. Through computational experiments, we confirmed the performance of the proposed algorithm in synthetic and real-world datasets.