Eager: Inferring Comprehensive Individual Traveler Information in Multi-Modal Travel Environment Using Automatic Fare Collection Data

This Smart and Connected Communities (S&CC) EArly-concept Grant for Exploratory Research (EAGER) project explores quantitative methods to infer the traits of individual traveler behavior (Origin-Destination trip information and traveler routes/mode choice preferences) from transaction-type transportation data in a multi-modal transit system. The inferential power of this project's approach relies on how the travelers revise their routing choices (recorded by transactions) in response to perturbations in the travel environment and its changing conditions. The inferred knowledge of travelers' origin-destination and preferences can be used to plan, monitor and predict the response of the travelers to operational decisions of public transit system managers and policy-makers, thereby increasing the system flexibility and operational efficiency. Such individual knowledge will be particularly useful for the planning and operations of alternative services in the events of transit system failures/closures, and also, further develop recently emerging customized transportation, including shared mobility systems and ride sharing systems, inform dynamic parking pricing, etc. As a creative educational activity effort, an on-campus test-bed will be created based on student data (student/employee card that is used across campus for various activities including bus, facilities access, dining, shopping, etc.), for various transportation informatics investigations. This research reaches into the yet untapped potential of Automatic Fare Collection data, more broadly transaction-type data, to inform Smart City and Transportation Informatics research. The project's methodological advances are general, i.e., not limited to any particular application. This research goes beyond the standard use of Automatic Fare Collection data use for identifying and understanding the statistical properties/trends, to distilling the hidden patterns of traveler agenda and behavior in multi-modal travel environment. Methodologically, the inference framework advances the Expectation Maximization (EM) paradigm that found much success in many inference tasks. The PIs suggest using Iterative EM and Selective Set EM methods, which promise to reliably infer two unknowns for each individual traveler: routing preference and Origin-Destination. If successful, the framework can start a new branch of methodological data-heavy research with repeated data, allowing for the studies of individual travel behavior at a finer granularity and presenting new opportunities to transportation policy makers.