¿µ¹®³»¿ë (English Abstract) |
Content popularity at the network edge is dynamic, i.e., the contents requested by users change over time. As a result, conventional content caching techniques make it difficult to guarantee Service Level Agreements (SLAs) and Quality-of-Experience (QoE) for seamless media streaming. Given the limited cache space and the high number of cache misses, a proactive content caching strategy is essential. Furthermore, in a distributed network infrastructure, the decision to cache or not cache the contents at different nodes affect the cache hit probability significantly. This paper proposes a distributed learning-based popularity prediction framework for proactive content caching that includes local prediction models to build a regional content popularity database. We use item-based collaborative filtering to construct a proactive content caching algorithm for seamless media streaming at the network edge. Simulation results show the efficiency of our proposed approach. |