Abstract: In order to improve the edge caching efficiency of Fog-Radio Access Network (F-RAN), a collaborative content caching strategy based on user preference prediction and content popularity prediction is proposed in this paper. Firstly, the Latent Dirichlet Allocation (LDA) model in the subject model is used to dynamically predict user preferences. Secondly, the topological relationships between different devices in the network and the obtained predicted user preferences are utilized to predict the changes of content popularity in an online fashion.Combining with correlation between the base station to reduce the repetition rate of cache content files. Finally, with the aim of maximizing the cache hit ratio, the Q-learning algorithm in reinforcement learning is used to obtain the optimal content caching strategy. Simulation results show that the proposed content caching strategy can effectively improve the cache hit ratio compared with existing content caching strategies.