AoI-Based Temporal Graph Attention Network for Content Update
Fan Jiang, Wei Wang, Lei Liu, Yongzhi Zhai, Haoyi Wang · 2024
In this paper, we present an AoI-based content prediction model utilizing a temporal graph attention network model to enhance cache hit rates and ensure content freshness. By capturing and aggregating time-based embedding features of both users and content, our model accurately predicts the popularity of requested content. Specifically, the concept of (Age of Information) AoI is incorporated to eliminate stale caching. Moreover, we also devise an optimal window for updating popular content, thereby markedly reducing user transmission latency and enhancing overall user experience quality. Experimental results demonstrate that our proposed AoI-based cache update strategy, combined with the temporal graph attention mechanism, significantly enhances cache hit rates and ensures the timeliness of cached content.