Estimating population density distribution from network-based mobile phone data

In this study we address the problem of leveraging mobile phone network-based data for the task of estimating population density distribution at pan-European level. The primary goal is to develop a methodological framework for the collection and processing of network-based data that can be plausibly applied across multiple Mobile Network Operators (MNOs). The proposed method exploits more extensive network topology information than is considered in most state-ofthe- art literature, i.e., (approximate) knowledge of cell coverage areas is assumed instead of merely cell tower locations. A distinguishing feature of the proposed methodology is the capability of taking as input a combination of cell-level and Location Area-level data, thus enabling the integration of data from Call Detail Records (CDR) with other network-based data sources, e.g., Visitor Location Register (VLR). Different scenarios are considered in terms of input data availability at individual MNOs (CDR only, VLR only, combinations of CDR and VLR) and for multi-MNO data fusion, and the relevant tradeoff dimensions are discussed. At the core of the proposed method lies a novel formulation of the population distribution estimation as a Maximum Likelihood estimation problem. The proposed estimation method is validated for consistency with artificially- generated data in a simplified simulation scenario. Final considerations are provided as input for a future pilot study validating the proposed methodology on real-world data.