An origin-destination level analysis on the competitiveness of bike-sharing to underground using explainable machine learning

Bike-sharing offers a convenient transportation option, enhancing the potential for direct competition with underground transportation, especially for short-distance trips. However, research on bike-sharing trips primarily focuses on survey data or aggregated data at the station-level. Few attempts have been made to under -stand the competition between bike-sharing and underground at the origin-destination (OD) level. This study aims to explore the competitiveness of bike-sharing to the underground at short-distance level using actual OD-level bike-sharing and underground ridership data collected in London. Light Gradient Boosting Machine and SHapley additive explanations models are employed for the analysis.Our results found that bike-sharing can serve as a competitor to the underground, especially in denser urban areas and peak periods. The competitiveness of bike-sharing is associated with the attributes of trips' origins and destinations, route characteristics, and time. In particular, the route characteristics of travel duration/distance, road gradient, bike infrastructure availability and the number of crossings are correlated with the competitiveness of bike-sharing to the underground. Moreover, it is found that users pay more attention to the characteristics of origins rather than destinations. Our findings can provide valuable implications for promoting bike-sharing as a substitution to underground service.