Abstract: Current activity-based modeling (ABM) has two major limitations. First, ABM conventionally relies on household travel survey (HTS) data, which suffers from low spatial heterogeneity due to a low sampling rate (e.g., 1-5% of the population) and a low collection frequency. This issue results in a poor spatial resolution of generated and forecasted travel demand. Second, ABM performs population synthesis and daily activity schedule generation independently due to the lack of data for spatiotemporal travel choices (i.e., activity time and destination choices). This causes a loss of interdependency between sociodemographic attributes and corresponding activity-travel choices. Given the continuous collection of mobility patterns at a high spatial resolution for a large proportion of the population from transit smart cards (SC), the fusion of HTS and SC data has the potential to address the above limitations. The talk presents a novel cluster-based data fusion method that exploits the benefits of both HTS and SC data to jointly generate a spatially heterogeneous synthetic population of individuals and their activity schedules. The properties of the proposed method are analytically derived to ensure an interpretable and trustworthy data fusion. The application of the proposed method is demonstrated using the HTS and SC data from Seoul, South Korea.
