Incremental Local Distribution-Based Clustering Using Bayesian Adaptive Resonance Theory

Most of the existing Bayesian clustering algorithms perform well on the balanced data. When the data are highly imbalanced, these Bayesian clustering algorithms tend to strongly favor the larger clusters, but provide a notably low detection of the smaller clusters. In this paper, we present an incremental local distribution-based clustering algorithm with the Bayesian adaptive resonance theory (ILBART). This algorithm is developed to adapt itself to a changing environment without using any predefined parameters. The algorithm not only accurately finds the clusters, even in data sets with a severely imbalanced distribution, but also efficiently processes the dynamic data according to the evolving relationships among the clusters. We test our proposed algorithm with experiments conducted on several imbalanced data sets. The experimental results show that our proposed algorithm performs well for clustering imbalanced data and can also obtain a better performance than many other relevant clustering algorithms in several performance indices.