Abstract

Traditional travel surveys, although grounded in solid sampling theory, are often costly and limited in their ability to accurately and comprehensively capture detailed travel behavior across space and time. GPS-based surveys, derived from mobile device location data (MDLD), offer complementary insights by tracking individuals’ movements and providing larger, more extensive samples of travel behavior. While these surveys have the potential to enhance our understanding of mobility patterns, they often lack essential demographic and socioeconomic information due to privacy constraints and sampling biases. This research proposes two approaches to integrate traditional travel surveys with GPS-based surveys using two record linkage algorithms: 1) A Two-Step Probabilistic Method, based on the Fellegi-Sunter framework, which evaluates trip similarity and uses a ranking algorithm to connect individuals. 2) A One-Step Fuzzy Matching Method that compares trip chains using Dynamic Time Warping (DTW) to generate similarity scores. Simulation results showed that both methods yield comparable F1 scores under low to moderate noise levels. However, the two-step probabilistic method demonstrates a significant drop in performance at higher noise levels, indicating its sensitivity to noise. In contrast, the fuzzy matching method maintains higher precision and overall robustness under the same conditions. To demonstrate the applicability of these approaches, we conducted a real-world case study by linking respondents from a Regional Travel Survey (RTS) to devices and their movements recorded in a GPS database. The findings highlight the potential of these methodologies to effectively integrate diverse data sources, thereby enriching the analysis and understanding of travel behavior.