Using mobile phone data analysis for the estimation of daily urban dynamics

The estimation of population dynamics has become a crucial public transport planning issue. The scope of this paper is the estimation of time variant population densities at fine-grained level using geolocalized mobile phone (MP) data. After preprocessing anonymized aggregated MP data of the complete Greater Paris area, we apply spatial mapping methods to project the MPs locations from network cells to census blocks. Prior to the calibration of MP densities with national census population (static model), we estimate blocks land-use to filter out noisy areas. Our loglinear regression model achieves high performance regarding several metrics, and our hybrid mapping method grants competitive performance with respect to the state of the art. Following our static parameters interpretation, we provide a novel relation for daily population dynamics. We validate this dynamic model with sport events attendances.