Learning-Bayesian network structure based on synergetics

Bayesian network is a probabilistic graphical model, which has been used for probabilistic reasoning in expert systems. Because that the novel method has a powerful ability for reasoning and a flexible mechanism to learning, it provides an effective way to deal with incomplete data or uncertainty. Learning Bayesian network from a given data set and prior knowledge is certainly the most difficult problem in the research domain. In this paper we give an introduction to Bayesian networks, and provide a review of previous works on learning Bayesian network structure. Then we propose a novel method for structure learning of Bayesian networks based on synergetics, it takes advantages of all the information include the given data set, expert knowledge and prior information, and trends to find the simplest structure which best fits the data. The experimental results prove that such an approach is feasible, robust and worth further analysis. Eventually, some remarks and some suggestions are given.