Learning hidden variables in Bayesian Networks with Bayesian Entropy Criterion for supervised classification

In this paper, we make use of a new criterion, the Bayesian Entropy Criterion (BEC), to learn hidden variable Bayesian Networks for supervised classification. This criterion takes into account the decisional purpose of a model by minimizing the integrated classification entropy. Experiments on real dataset show that BEC performs better than the BIC criterion to select a model minimizing the classification error rate. Learning hidden variable structures with BEC, we can find the more effective hidden variables for supervised classification model, which may reveal some valuable principles of certain domain.