Mobile traffic data as a proxy for urban mobility: a preliminary study in Paris

As 4G and 5G networks become the backbone of urban infrastructure, the granular mobile traffic data they generate offers a high-resolution lens into collective human mobility dynamics. This paper investigates the hypothesis that service-specific mobile traffic can serve as a reliable proxy for human mobility and urban behavior. We model the problem as a Time Series Anomaly Detection (TSAD) task by proposing an unsupervised framework that employs a reconstruction-based representation model to establish a baseline urban activity. Our pipeline identifies significant deviations in social media app usage (e.g., Twitter, Instagram, YouTube) as spatio-temporal anomalies. These anomalies are subsequently clustered to filter noise and reveal coherent patterns corresponding to real-world events. We validate our framework using the NetMob 2023 dataset across the Paris metropolitan area. Our preliminary results demonstrate that this approach effectively captures a diverse spectrum of mobility events, including political protests, sporting matches, and disasters. These findings suggest that service-specific traffic data is a powerful, non-intrusive indicator of urban mobility flows and social disruptions.