The past communication behaviors that guide the system communication in the future to satisfy the requirements of users and adapt to the changes of environment are the core part of cognitive radio system. In this paper, a cognitive engine based on Bayesian network is proposed to solve the parameters self-adaptive-adjusting problem of cognitive radio system under the complicated and highly varying radio environment and user requirement. Through structure learning and parameter learning of the sample data from the past communication behaviors, cognitive engine is established. The states of radio environment and requirements of users are made as inference evidences by data preprocessing, and the cognitive engine is used to make decision of the configuration parameters of communication system, and then the reconfiguration system is completed. A mobile wireless network is modeled to finish reconfiguration simulation using OPNET tool in this paper. Simulation results show that the proposed cognitive engine can make the wireless mobile network adapt to environment and effectively improve end-to-end communication performance. The feasibility of the method to model cognitive engine with Bayesian network is validated in this paper.