A Diversified Recommendation Scheme for Wireless Content Caching Networks

Wireless cellular networks currently face constantly growing data demands, which lead to network congestion and high latency. Cache-aware recommendation can reshape users' request behavior and improve cache efficiency in wireless caching systems, thus alleviating network congestion and shorting transmission delay. However, existing cache-aware recommendations serve the caching system by reducing the quality of recommendations. Specifically, it usually provides only limited and similar content items and lacks diversified recommendation services, which severely reduces users' satisfaction. To tackle this challenge, we propose a diversified recommendation mechanism-based solution that aims to simultaneously improve the performance of wireless caching systems and the quality of recommendation to improve users' satisfaction to a greater extent. To this end, we propose a quantitative model that captures the impact of recommendation decisions on the diversity of recommendation sets. This model enables us to formulate a joint cache hit ratio and recommendation diversity maximization problem, taking into account each user's recommendation size and cache capacity requirements. Since this problem is a nonconvex integer programming problem, we decompose it into two subproblems, i.e., the cache placement problem and the diversified recommendation problem. Then we design Tabu search-assisted and simulated annealing-oriented algorithms to solve these two subproblems, respectively, and perform iterative alternating optimization for the whole problem. Monte-Carlo simulation validates the effectiveness of our method in terms of cache hit ratio and recommendation diversity compared to various benchmarks.