Oke, JimiMohammed, Mohammed A.2024-10-022024-10-022024-0510.7275/54856https://hdl.handle.net/20.500.14394/54856Regional transit systems often lack the resources for comprehensive data collection systems such as smart cards, which hinders their ability to understand passenger travel patterns and optimize service. This thesis addresses this gap by developing an innovative trip chaining framework that uses boarding-only data from mobile ticketing systems, a more accessible and cost-effective alternative. The framework incorporates passenger typology, seasonality, and spatial error correction to address the unique challenges posed by this data. First, spatiotemporal passenger types are identified through clustering analysis of mobile ticketing activations, revealing distinct travel behaviors and patterns. These types inform the calibration of type-specific and season-aware parameters for boarding, alighting, and transfer inferences within the trip chaining framework. Additionally, a gradient boosting machine model is trained to learn the spatial error structure of initial predictions and refine the model's accuracy. The results demonstrate that the integration of passenger typology, seasonality, and spatial error correction significantly enhances the accuracy of trip chaining predictions. Notably, incorporating spatial error correction through gradient boosting leads to a substantial improvement in model performance, with a 70% reduction in mean absolute error. Furthermore, the passenger typology offers valuable insights into travel characteristics of different user groups, facilitating more targeted planning efforts. This framework provides regional transit planners with a powerful tool to extract valuable insights from limited data sources. By understanding passenger travel patterns and demand, transit agencies can optimize service, allocate resources efficiently, and ultimately promote greater public transportation usage.Trip chainingTypology analysisMachine learningPublic transitSpatiotemporal patternsMobile ticketingThesis (Open Access)https://orcid.org/0000-0003-4660-8030