Why metro passengers change travel behavior: Individual-level insights from interpretable machine learning

Document Type

Journal Article

Publication Date

2025

Subject Area

place - urban, mode - subway/metro, technology - ticketing systems, technology - passenger information, ridership - behaviour, ridership - demand, ridership - forecasting

Keywords

Metro, travel behavior, smart card data, demand forecasting

Abstract

Understanding individual-level travel behavior changes is essential for tailoring metro services to meet dynamic passenger needs, yet existing research focuses on aggregate trends, ignoring micro-level adaptation mechanisms. This study bridges this gap through an innovative, disaggregated framework combining smart card data, machine learning, and interpretable AI to decode drivers of metro passenger behavior changes. Utilizing smart card data from Shanghai's sixteen-line metro network, we reconstruct daily travel diaries across over ten million passengers and classify for types (i.e., regular patterns, time changes, route changes, and trip cancellations) using an XGBoost model. SHAP value analysis reveals holidays as systemic disruptors, reducing regularity and increasing cancellations, and trip frequency as a dualistic driver, route adaptability for high-frequency travelers but temporal rigidity. Weekly frequency volatility also strengthens cancellations, while weather factors show minimal impact. Our framework empowers metro operators to move beyond reactive crowd control toward proactive, personalized management. By quantifying how specific factors influence behavior, this approach enhances demand forecasting and opens the door to more tailored services. This marks a significant step forward in adapting metro systems to the diverse and dynamic needs of passengers.

Rights

Permission to publish the abstract has been given by Elsevier, copyright remains with them.

Link to Full Text

Share

COinS