IGMM-CD: A Gaussian Mixture Classification Algorithm for Data Streams with Concept Drifts

Abstract

Learning concepts from data streams differs significantly from traditional batch learning, because in data streams the concepts to be learned may evolve over time. Incremental learning paradigm is a promising approach for learning in a data stream setting. However, in the presence of concept drifts, outdated concepts can cause misclassifications. Although several incremental Gaussian mixture models methods have been proposed in the literature, we notice that these algorithms lack an explicit policy to discard outdated concepts. In this paper, we propose a new incremental algorithm for data stream learning based on Gaussian Mixture Models. The proposed method is compared to various algorithms widely used in the literature, and the results show that it is competitive with them in various scenarios, overcoming them in some cases.