Joint Pushing and Recommendation for Susceptible Users with Time-Varying Connectivity

Proactive caching holds the promise of increasing network throughput by caching popular contents at mobile users. The cache hit ratio heavily relies on the demand probability of a user for a content. As a powerful tool to shape user demand, the recommendation system has been introduced into caching networks to further improve the cache hit ratio. However, the joint design of pushing and recommendation for users with time-varying connectivity is still unknown. In this paper, we study joint pushing and recommendation (JPR) policies that maximize the effective throughput in wireless networks with time-varying user connectivity. Two offline and one online JPR policies are presented based on noncausal, statistical, and causal connectivity information, respectively. It is shown that the causal feedback brings significant throughput gain, especially when the recommendation weight and the connection probability are relatively small.