University courses are not isolated settings. There is a certain link among them. This paper presents a construction method for Bayesian networks of university courses relationship, which use the examination results of students' courses as data sample. The undirected graph was constructed with structure learning algorithm based on information theory, and its edges were oriented according to the time order of courses opening. Such the Bayesian network of courses dependence relationship was obtained, and its condition probability table was learned by mathematical statistics method. This model had represented the dependence relationship of courses intuitively, and the condition probability table quantified tightness of relationship. It plays guidance role to the setting and arrangement of university courses, and it has the prediction ability to the studentspsila achievement.