Spatial Analysis of Mobile Traffic Hot Spots

An enormous increase in data traffic demanded by mobile users calls for effective (and efficient) deployment strategies such as multi-layer heterogeneous networks or ultradense small cell networks. However, placing small cells at the desired locations to offload as much traffic as possible from overlaying macro cells is a crucial task. Geo-location and user equipment positioning techniques, e. g. those based on so-called call traces, help obtain spatial distributions of user locations and their respective traffic volumes. In this paper, we provide a tool capable of reducing errors that stem from spatial discretization of traffic data and that autonomously detects hot spots and larger hot zones given a certain threshold. Based on geo-located traffic volumes in a 3G network in a dense urban European city, we find that traffic is approximately log-normally distributed in the area and that the size of traffic hot spots (where traffic demand exceeds a certain threshold and with an area which can be covered by individual small cells) are approximately Weibull distributed. Based on our statistical findings, we observe that deployment of 4 small cells per km2 covering 3.2 % of the total area and around 34 % of the total traffic volume is a very meaningful strategy; however, deploying more small cells in hot zones becomes increasingly costly in terms of the ratio of area covered and traffic demand serviced.