Abstract: Bayesian network structural learning plays a very important role in the processing of Bayesian network’s construction, and an effective structural learning algorithm is the base of constructing the optimum Bayesian network. An algorithm of Bayesian network structural learning(called MIBNS) based on mutual information is proposed. The algorithm can give the concealed dependency relationships among data attributes, and make dimension reduction at the right moment, which can improve the performed efficiency and ensure the accuracy rate. Experimental result shows that the algorithm is effective. Compared with the SGS, the algorithm of MIBNS is more effective in the similar results.