Predicting User Mobility in Mobile Radio Networks to Proactively Anticipate Traffic Hotspots

Abstract

With approx. 6 million macro cells worldwide and a gross energy consumption of approx. 100 TWh per year as of 2013, mobile networks are one of the major energy consumers in the ICT sector. As trends, such as cloud-based services and other traffic-intensive mobile applications, fuel the growth of mobile traffic demands, operators of mobile telephony networks are forced to continuously extend the capacity of the existing infrastructure by both implementing new technologies as well as by installing new cell towers to provide more bandwidth for mobile users and improve the network's coverage. In order to implement energy-efficient reconfiguration mechanisms in mobile telephony networks as proposed by the project Communicate Green, it is essential to anticipate traffic hotspots, so that a network's configuration can be adjusted in time accordingly. Hence, predicting the movement of mobile users on a cellular level of the mobile network is a crucial task. In this paper, we propose a Movement Prediction System based on the algorithm of Yavas et al. that allows to determine the future movement of a user on a cellular level using precomputed movement patterns. We extended the algorithm to be capable of preselecting patterns based on time and contextual data in order to improve prediction accuracy. The performance of the algorithm is evaluated based on real and live user movement data from the OpenMobileNetwork, which is a platform providing estimated mobile network topology data. We found that the algorithm's prediction quality is sufficient, but requires an extensive amount of recorded user movements to perform well.