Identifying Common Periodicities in Mobile Service Demands with Spectral Analysis

Abstract

In this paper, we investigate the existence and prevalence of comparable dynamics in the temporal fluctuations for the traffic demands generated by mobile applications. To this end, we hinge upon a spectral analysis framework, by computing Discrete Fourier Transforms of the typical demands for tens of popular mobile services observed in an operational metropolitan-scale network. We filter, cluster, and analyse hundreds of frequency components, and identify a substantial set of regular patterns that are common across most service demands. We also unveil how several mobile services defy classification, and have instead highly distinguishing temporal dynamics.