Differences in bike-sharing usage and its associations with station-surrounding characteristics : a multi-group analysis using machine learning techniques

Abstract

The bike-sharing system offers a wide range of benefits to promote human mobility for all. However, many bike-sharing systems are most used by specific demographic groups (e.g., younger people and males), suggesting that the resulting benefits are not equally distributed among the public. We aim to empirically examine the differences in bike-sharing usage among varying demographic groups and its association with station-surrounding characteristics (i.e., land use, transportation infrastructure, and population distribution) in Tartu (Estonia) using a machine learning approach (i.e., gradient boosting decision trees). The results revealed that the floor area ratio played an extremely important role in promoting bike-sharing usage, but such a strong positive impact was not observed within senior groups. Instead, bike-sharing usage by seniors was strongly positively associated with the commercial land and bike lanes. It also detected that male teenagers and young adults were less likely to be influenced by the public land than their female counterparts when using shared bikes. Shared bikes located in areas with dense male senior residents gained high usage by them; however, such phenomenon was not observed from their female counterparts. These findings can provide significant insights for interventions targeting demographic-specific bike-sharing usage to promote inclusivity and equity in urban transportation.