Mohit Talreja, Joshua Diao, Jim James, Radu Casapu, Tejas Santanam, Ethan Mendes, Alan Ritter, Wei Xu, James Hays. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Many special events, including sports games and concerts, often cause surges in demand and congestion for transit systems. Therefore, it is important for transit providers to understand their impact on disruptions, delays, and fare revenues. Ridership after large sporting events is distinct from many other ridership patterns due to the high density of ridership localized to a few nearby stations. This paper provides the following novel methodology for long-term planning of large-event, post-game ridership by 1) predicting the total post-game ridership; 2) combining the total prediction with historical trends to forecast the passenger flow curve at nearby stations after the game; and 3) estimating the required train frequencies to serve these customers with minimal passengers left behind by each train. Additionally, this paper proposes a suite of data-driven techniques that together create a data-driven pipeline to exploit Automated Fare Collection (AFC) data for evaluating, anticipating, and managing the performance of transit systems. This paper includes a case study where the proposed pipeline is used to generate an adjusted train schedule for the post-game period and simulated with the rail ridership data from the Metropolitan Atlanta Rapid Transit Authority (MARTA). The simulation results highlight how the proposed schedules based on the estimated required post-game train frequencies could significantly improve post-game congestion and wait time. Furthermore, the results show that the long-term post-game demand forecasts could be an effective tool for tactical planning decisions such as the number of additional trains and operators that are needed during post-game periods compared to the regularly scheduled timetables.