Web services bring more conveniences for users and developers. However, it makes user face the problem of service information explosion. The personalized service recommendation solves the problem. This paper proposes a mechanism to evaluate Web service composition’s reliability which bases on user context information in mobile computing. The mechanism constructs user behavior model by formatting user context and then quantitative verification is performed to estimating whether the recommended Web service composition satisfy the requirements of users. Finally, experiments are carried out to demonstrate the effectiveness of our mechanism.