Temporal-Spatial Mobile Application Usage Understanding and Popularity Prediction for Edge Caching

Abstract

The explosive growth of smart devices and expansion of network services drives the flourish of mobile applications. Caching popular application services at the edge, BSs, or APs, which are closer to end users, could significantly improve the user experience and network capacity. To exploit this potential, it is critical to understand the traffic consumption and app usage pattern under BSs in the metropolitan area. Mobile big data collected from network interfaces facilitates the data-driven approach in characterizing these features. This article aims to design an edge caching strategy for app services based on the observed characteristics of BSs in terms of points of interest (POIs), logs, and traffic generated by various categories of apps. We first analyze the temporal characteristics of different categories of apps, and then further investigate the logs and traffic generated by app types under different BSs clustered by POIs. The top N popular app types in a given period of time under different BS clusters are predicted, which is helpful for network operators to know the traffic distribution of different app services over all BSs and design the edge caching scheme.