Learning-to-Rank Based Strategy for Caching in Wireless Small Cell Networks

Abstract

Caching in wireless network is an effective method to reduce the load of backhaul link. In this paper, we studied the problem of wireless small cell network caching when the content popularity is unknown. We consider the wireless small cell network caching problem as a ranking problem and propose a learning-to-rank based caching strategy. In this strategy, we use the historical request records to learn the rank of content popularity and decide what to cache. First, we use historical request records to cluster the small base stations (SBS) through the k-means algorithm. Then the loss function is set up in each cluster, the gradient descent algorithm is used to minimize the loss function. Finally we can get the ranking order of the content popularity for each SBS, and the files are cached to the SBS in sequence according to the order. From Simulation results we can see that our strategy can effectively learn the ranking of content popularity, and obtain higher cache hit rate compared to the reference strategies.