Mobile Data Access has been an increasingly popular service along with the development of mobile technologies. Many research problems depend on the details of access requests when studying data access in the mobile environment. If we can get a whole picture of data service requests, it is possible to design a proper model and service mechanism according to the global situation. In this paper, we propose a user pattern analysis approach to predict the global user requests in cellular systems, which is composed of two parts: user movement pattern analysis and user access pattern analysis. Simulation results show that our approach achieves a good performance with an acceptable computational complexity.