Data Fusion for Travel Analysis: Linking Travel Survey and Mobile Device Location Data

Abstract

Abstract Travel surveys typically collect detailed information about demographics and travel behavior of households and persons; but their sample sizes are often limited, and trip information is usually limited to a single day. In contrast, Mobile Device Location Data (MDLD) provides extensive and accurate trip records spanning multiple days for each person from a much larger sample, while demographic information for the individuals are always lacking due to anonymization. This study constructs data panels combining high-precision, long-term trip records from MDLD with detailed demographic information from a regional travel survey (RTS). Two probabilistic record linkage algorithms are employed to identify individuals with similar travel behaviors between RTS and MDLD datasets. The data panels constructed by the linkage algorithm captured not only peak-hour commutes but also off-peak travel and non-home-related trips, shedding light on previously underreported travel behaviors and offering a more holistic view of individuals' travel patterns. This comprehensive dataset also exhibits comparable demographic characteristics to the original RTS, showing that such data panel is a reasonable representation of the entire population. The integration of diverse datasets holds promise for revolutionizing travel behavior analysis and shaping the future of transportation planning in the era of mobile technology and big data.