Cluster-based content caching driven by popularity prediction

This paper introduces a novel cache replacement strategy, named Cluster-Based Content Caching (CBCC). CBCC extracts the content feature and uses it to predict popularity for cache replacement. The prediction of expected popularity of new content is based on the clustering algorithm dynamically. If the content is not in the cache list and the list is full, CBCC can help the Cache Node (CN) to make smarter decisions about whether cache new content or not and which content will be replaced based on a popularity prior queue, intending to increase the number of cache hitting in the near future. This strategy can reduce the pressure on the backbone and further improve user satisfaction. We evaluate the performance of our algorithm based on an open Douban Movie Comment Dataset. This dataset can be used to simulate users’ requests. Simulation results show the feasibility and effectiveness of the proposed algorithm.