Three-way decisions based Bayesian Network

Rough set theory provides a ternary classification method by approximating a set into positive, boundary and negative regions. A Bayesian Network classifier is a directed acyclic graph model that encodes a joint probability distribution over a set of random variables. In this paper, we propose a probabilistic rough set model, three-way decisions based Bayesian Network (3DBN), to integrate these two classification techniques. Several comparison experiments are implemented to evaluate the performance of 3DBN. Experimental results show that the proposed method can get a better performance on accuracy and misclassification cost.