Minimum cost attribute reduction in three-way decisions based Bayesian network

Three-way decisions based Bayesian network is an integrated model by combining three-way decisions theory and Bayesian network. Compared to classical two-way decisions based Bayesian network, the three-way decisions based Bayesian network could obtain a lower misclassification error. Another advantage of three-way decisions based model is that it can generate minimal decision cost. Based on this, a new minimum cost attribute reduct in three-way decisions based Bayesian network is defined, and a heuristic attribute reduction method for the reduct definition is designed in this paper. Based on the proposed attribute reduct, one can get the final classification result with minimal decision cost. Experimental results on several data sets show the efficiency of the proposed attribute reduct method.