How to improve the user experience for the networks is a problem should be extensively concerned about. In this paper, an approach to networks congestion analysis model based on Bayesian Network is proposed to realize the network system operation and maintenance effectively. Compare to the networks topology, the congestion pattern networks is constructed and each congestion object's conditional probability table and the joint probability distribution can be calculated according to the network data flow. The dependency between the congestion nodes can be presented and we can find the abnormal point in the networks topology. Experiment is conducted to verify the correctness and validity of the proposed method, and the results indicate that it is effective to construct the congestion pattern model with Bayesian Network, and it can be able to dig out the potential congestion path more accurately.