In this paper, through the analysis of the characteristics of particle swarm optimization algorithm, combined with the specific circumstances of Bayesian network structure learning, proposed to based on improved particle swarm algorithm.The algorithm uses the BIC measure function as a standard Bayesian network, while preserving the optimal particle case, the possibility of a mutation operation is added to decrease the algorithm into a local optimum. Through a typical Asia network, show that the algorithm is feasible, and other related algorithm is better than the experiment, the effectiveness of the algorithm. In this paper, the algorithm is verified from two aspects of theory and experiments, the results show that the algorithm is feasible.