Semi-supervised Bayesian ensemble method for concept drift

Abstract

This study focuses on classification problems in domain adaptation, with a specific emphasis on addressing concept drifts in streaming data. To enhance classification accuracy, we propose a novel semi-supervised ensemble algorithm based on a Bayesian constrained ridge model. Our approach employs the Gibbs sampling algorithm to estimate the weights of the ensemble model and infer the prediction distribution. By utilizing this distribution, we incorporate highly confident predicted labels into the learning process and update the ensemble weights, thereby improving classification performance. To validate the effectiveness of our method, we conduct extensive experiments on synthetic and real-world datasets, comparing it with state-of-the-art semi-supervised algorithms. Our results demonstrate the superior performance of the proposed approach.