Multi-Head Attention Based Popularity Prediction Caching in Social Content-Centric Networking With Mobile Edge Computing
Jie Liang, Dali Zhu, Haitao Liu, Heng Ping, Ting Li, Hangsheng Zhang, Liru Geng, Yinlong Liu
Abstract
With the rapid growth of social network traffic, the design of an efficient caching strategy is crucial in the social content-centric network (SocialCCN). In order to design a more comprehensive popularity prediction caching strategy, in this letter, we proposed a novel architecture that integrates mobile edge computing (MEC) in SocialCCN (MeSoCCN) and proposed multi-head attention based popularity prediction caching strategy in MeSoCCN. Firstly, we proposed a multi-head attention based popularity prediction model (MAPP) that considers multi-dimensional features including history and future popularity, social relationships, and geographic location to predict content popularity. Then, we design a caching strategy based on the prediction results of MAPP. The simulation results show that the proposed MAPP model achieves lower predictive error and the proposed predictive caching strategy improves cache hit rate and reduces hop redundancy in the network.