Vehicular traffic predictions from cellular network data — A real world case study

The emergence of mobile technologies provides the opportunity to carry mobility field into the smart city arena. Transportation data are key factors for improving mobility services: traditional approaches to compute urban dynamics, mobility patterns and real-time vehicular traffic situations are based on cameras, on-road sensors or emergency calls, while more modern approaches merge social warnings and mobile data in collaborative navigation systems to detect traffic congestions. In this paper, we present a novel “passive” approach for gathering, processing, and predicting real-time vehicular traffic conditions from cellular network data. The approach exploits the regression statistical tool to detect whether significant statistical models exist to describe correlations between cellular network events and vehicular traffic situations. The paper discusses the regression model we derived and it presents the results obtained by validating our approach, in a real industrial setting and for the Milan city, against the well-known traffic solutions: Autostrade.it, InfoBlu and Google.