Spatiotemporal travel behavior modeling and analysis for better public transport systems
The public transportation system is crucial in alleviating urban congestion. The widespread of smart card automated fare collection (AFC) system produces massive data recording passengers’ day-to-day transport dynamic, which provides unprecedented opportunities to researchers and practitioners to understand and improve transit services. This project aims to make full use of the transit operational data (mainly smart card data) to enhance transit services. The main body of the research project is spatiotemporal behavior patterns mining. The project collaborates with the transit operator exo and will be accomplished by a series of methodological and practical contributions. Data fusion technique will be used to make up incomplete data and contextualize trips’ and passengers’ attributes. We will construct features to profile passengers’ behavior at different time scales and utilize statistical learning methods to extract meaningful latent representations to help to understand passenger behavior patterns. The long-term changes in behavior patterns will be studied. Three sub-problems will be investigated together with the main project: 1) quantifying transit usage and operational service level, 2) transit fare scheme evaluation, and 3) interactions between multiple transportation modes.