Proactive Edge Caching With Popularity Prediction and Content Replication

The increasing demand for low-latency content delivery in 5G networks highlights the need for intelligent edge caching solutions. This paper proposes a cooperative edge caching framework for small-cell base stations (SBSs), leveraging overlapping cache assignments and a dynamic programming-based (DP) optimization algorithm to minimize content retrieval delay and maximize cache hit ratio (CHR). The algorithm incorporates both global and local content popularity predictions—obtained via exponential moving average (EMA) and LSTM models—and dynamically scales content placement decisions based on delay-aware replication benefits. Simulation results, conducted under diverse and realistic content demand patterns, demonstrate that the proposed approach significantly outperforms both traditional caching methods, such as Least Recently Used (LRU) and random assignment, as well as state-of-the-art demand-aware algorithms in terms of average delay and cache efficiency. The framework is well-suited for dynamic mobile edge computing (MEC) environments and provides a scalable foundation for next-generation cooperative caching strategies.