K-Dependence Bayesian Classifier Ensemble

Abstract

To maximize the benefit that can be derived from the information implicit in big data, ensemble methods generate multiple models with sufficient diversity through randomization or perturbation. A k-dependence Bayesian classifier (KDB) is a highly scalable learning algorithm with excellent time and space complexity, along with high expressivity. This paper introduces a new ensemble approach of KDBs, a k-dependence forest (KDF), which induces a specific attribute order and conditional dependencies between attributes for each subclassifier. We demonstrate that these subclassifiers are diverse and complementary. Our extensive experimental evaluation on 40 datasets reveals that this ensemble method achieves better classification performance than state-of-the-art out-of-core ensemble learners such as the AODE (averaged one-dependence estimator) and averaged tree-augmented naive Bayes (ATAN).

Keywords

• k-dependence forest
• diversity
• conditional dependencies