Caching Improvement Using Adaptive User Clustering

TL;DR

This paper proposes an adaptive user clustering method for proactive caching in 5G small cell networks, aiming to improve cache efficiency by tailoring content to user groups based on behavior, with an analytical model and optimization strategy.

Contribution

It introduces a novel user clustering approach for caching, including a method to estimate cluster count and an analytical expression for hit probability.

Findings

Cluster-based caching improves hit probability.

Akaike information criterion effectively estimates cluster number.

Optimized base station association enhances cache performance.

Abstract

In this article we explore one of the most promising technologies for 5G wireless networks using an underlay small cell network, namely proactive caching. Using the increase in storage technologies and through studying the users behavior, peak traffic can be reduced through proactive caching of the content that is most probable to be requested. We propose a new method, in which, instead of caching the most popular content, the users within the network are clustered according to their content popularity and the caching is done accordingly. We present also a method for estimating the number of clusters within the network based on the Akaike information criterion. We analytically derive a closed form expression of the hit probability and we propose an optimization problem in which the small base stations association with clusters is optimized.