Location-Aware Caching via Predicting Heterogeneous File Preferences in Mobile Networks

Abstract

We present a new caching method in content-centric networks (CCNs) where each mobile user is equipped with finite-size cache and device-to-device content delivery is employed. In this study, we exploit the heterogeneity in file preferences among users who are moving around different locations (e.g., points-of-interest) in caching. Moreover, to infer our model parameter more accurately, we apply data imputation, which is a technique to replace unknown data with estimated values, based on collaborative filtering (CF). Our experimental results with real-world datasets demonstrate the superiority of our method over benchmark caching methods utilizing no location information in terms of both average hit ratio and runtime complexity.