Bayesian network is a graphical model that encodes probabilistic relationships among variables of interest. It is a natural way to express the causal information, and to discover the hidden patterns among the data. Learning of Bayesian network is to find out a network model that best represents the dependent relationships of the variables in a database, that is, given sample D and prior knowledge ζ, to find a Bayesian network S that fits the maximum posterior probability p(s^h|D,ζ). In this paper, the learning process of the network is strictly derived, and a case study is presented to indicate the applications of Bayesian network in data mining.