Residency and worker status identification based on mobile device location data

Abstract

Mobile device location data (MDLD) have been widely recognized for their rich human mobility information and thus considered as a supplementary data source for the current travel data bank consisting of travel survey data and traffic monitoring data. However, the lack of ground truth information about the device owners raises concern about the biases and representativeness of the nonprobability MDLD sample and significantly limits the applications of MDLD. This paper focuses on identifying two important socio-demographic characteristics for the MDLD sample devices: residency and worker status, including four worker types (normal commuters, professional drivers, mobility-for-work workers, and nonworkers/home-based workers). Based on the spatial–temporal patterns of location sightings and derived trips from MDLD, a comprehensive imputation framework is proposed with parameters calibrated against public domain ground truth data. A national-level case study in the U.S. based on a commercial MDLD dataset further evaluates the performances of the proposed heuristic rules. The multi-level validation results indicate a satisfying match against the ground truth data and prove the effectiveness of the proposed methods. As one of the earliest efforts to identify the residency and worker status information for a large-scale national-level MDLD dataset, mobile workers—including professional drivers and mobility-for-work workers—are also identified from MDLD for the first time.