Recommendation-Enabled Caching Strategy With Age of Information in Edge-Cloud Networks

The recommendation of local content in edge servers is a crucial strategy to alleviate cloud pressure and enhance resource utilization in edge-cloud networks. However, the dynamic complexities of user request behavior and the diminishing value of cached content have not been adequately accounted for in time-varying recommendation-enabled caching systems. In this study, an edge caching problem integrating content recommendation and Age of Information (AoI) based on user requests and content value is proposed to maximize system revenue in time-varying scenarios. Firstly, a personalized user request model is proposed to capture dynamic influences from content recommendation and user historical requests. Then, the optimization problem is decomposed into two subproblems: the content recommendation problem and the cache replacement problem. Moreover, the content recommendation problem is proved to exhibit monotone submodularity, and a greedy algorithm with $(1 - \frac{1}{e})$-approximate solution is proposed to determine the recommendation set for each user. Furthermore, the cache replacement problem is formulated as a Markov decision process, and an iterative optimization algorithm combining Recommendation and AoI based on Double Deep Q-Network (RA-DDQN) is proposed to maximize the long-term system revenue. Finally, extensive experiments conducted on a real dataset validate the superiority of the proposed algorithm compared to other algorithms.