Deep Regression Model for Videos Popularity Prediction in Mobile Edge Caching Networks

In recent years, the wide spread adoption of mobile and multimedia applications has resulted in exponentially increasing multimedia traffic, which exerts a great burden on backhaul links and mobile core networks. Mobile edge computing (MEC) alleviates the problem by enabling mobile edge devices with cache storage and allowing them to store popular multimedia contents requested by users to reduce network congestion and content delivery latency. However, to decide the multimedia contents to cache in the edge devices, the popularity of contents needs to be taken into consideration. In this paper, we propose a deep regression-based video popularity estimation for proactive video caching in MEC networks. In each time slot, an edge device, i.e., base station (BS) generates local estimates on content popularity, which are then shared by the neighboring edge devices. Then, each edge device performs popularity prediction using a deep regression technique for proactive content caching for the next time slot. Simulation results show that the proposed deep regression based method for videos popularity prediction achieves good performance and reduces latency significantly.