Learning Bayesian networks from incomplete databases using a novel evolutionary algorithm

This paper proposes a novel method for learning Bayesian networks from incomplete databases in the presence of missing values, which combines an evolutionary algorithm with the traditional Expectation Maximization (EM) algorithm. A data completing procedure is presented for learning and evaluating the candidate networks. Moreover, a strategy is introduced to obtain better initial networks to facilitate the method. The new method can also overcome the problem of getting stuck in sub-optimal solutions which occurs in most existing learning algorithms. The experimental results on the databases generated from several benchmark networks illustrate that the new method has better performance than some state-of-the-art algorithms. We also apply the method to a data mining problem and compare the performance of the discovered Bayesian networks with the models generated by other learning algorithms. The results demonstrate that our method outperforms other algorithms.