A Hierarchical Approach for Investigating Social Features of a City from Mobile Phone Call Detail Records

Abstract

Cellphone service-providers continuously collect Call Detail Records (CDR) as a usage log containing spatio-temporal traces of phone users. We proposed a multi-layered hierarchical analytical model for large spatio-temporal datasets and applied that for the progressive exploration of social features of a city, e.g., social activities, relationships, and groups, from CDR. This approach utilizes CDR as the preliminary input for the initial layer, and analytical results from consecutive layers are added to the knowledge-base to be used in the subsequent layers to explore more detailed social features. Each subsequent layer uses the results from previous layers, facilitating the discovery of more in-depth social features not predictable in a single-layered approach using only raw CDR. This model starts with exploring aggregated overviews of the social features and gradually focuses on comprehensive details of social relationships and groups, which facilitates a novel approach for investigating CDR datasets for the progressive exploration of social features in a densely-populated city.