Modeling Purposes of Public Transportation Trips for Human Need-Responsive Urban Mobility Efficiency

Public transportation (PT) systems are the artery systems for urban residents to access resources essential for fulfilling their daily needs. Enhancing the operational efficiency of PT systems is thus of critical importance in urban mobility improvement and sustainable city development. However, current PT system operations do not account for the impact of operation decisions on the satisfaction of diverse needs of riders, irresponsive to the fundamental human needs driving mobility behaviors of PT riders. To address this limitation, it is imperative to model and infer the purposes of PT trips to understand the types of human needs that these trips aim to satisfy. As such, this paper presents a Bayesian probabilistic method for station-level, hourly PT trip purpose modeling and inference. The proposed method integrates (1) the gravity model for modeling spatial trip purpose distributions and (2) a temporal variation model for modeling temporal trip purpose distributions to enable station-level, hourly PT trip purpose modeling and inference. The method was validated by comparing the modeled PT trip purpose distributions to those obtained from established mobility surveys. The validation results showed a strong alignment of the two types of distributions, with a Kullback-Leibler divergence score of 0.061 and a Jensen-Shannon divergence score of 0.014. Building upon the validation, this paper further implemented and demonstrated the proposed method in modeling and analyzing shifts in PT trip purposes during the COVID-19 pandemic in New York City. Temporal-spatial analysis revealed distinct patterns in the trip purpose shifts across time and space.