Uncovering Latent Patterns in Service-Level Spatiotemporal Mobile Traffic

Abstract

As people increasingly rely on mobile applications for their daily activities and needs, the digital traces left by smartphones have become an essential instrument to provide information about human activities. In this work, we focus on identifying latent structures in mobile traffic by targeting the traffic generated by individual applications, whose correlation with the urban environment has yet to be explored. Therefore, our mobile traffic analysis considers a third application-related dimension in addition to traditional spatial and temporal ones. In this context, we show that tensor decomposition techniques can be applied to service-level mobile traffic and improve the identification and interpretation of patterns that may remain undiscovered if conventional approaches are adopted.