News- Excited to share our latest publication on Transportation Research Part E

Our paper titled “Simultaneous Estimation of Induced, Diverted, and Ex-Post Demand for Railway Passengers: An Interpretable Machine Learning Framework Based on Constrained Computational Graphs” is now officially published and available online:

🔗 https://lnkd.in/ehF7ZhE4

In transit and railway planning, it is crucial to understand how total passenger demand is composed and influenced by different behavioral responses:

Diverted Demand: Shifts in route choices due to alternative line plans

Induced Demand: Increased ridership resulting from service improvements

Ex-Post Demand: Adjustments caused by capacity constraints like seat availability

To address this, we developed a constrained computational graph (CG)-based learning architecture, offering an interpretable and consistent machine learning framework that jointly estimates all three demand components and their behavioral sensitivities to service factors such as time, fare, and frequency.

This work offers practical insights for demand forecasting, network optimization, and transit policy design. Grateful to all collaborators who made this work possible!