This paper applies the reinforcement learning models which put psychology and economics into studying science and technology user learning behavior of information searching methods. Firstly, two traditional reinforcement learning models of Bush-Mosteller model and Börgers-Sarin model are introduced. Then, based on analysis of the reinforcement learning mechanism, control experiment is designed to explore user learning behavior through quality study. The experiment result indicates that reinforcement learning could successfully reflect the process of user learning and Börgers-Sarin model has got better fitting and forecasting effect. Finally, based on the experiment result, this paper provides some customer strategies for academic database website.