How the built environment affects E-scooter sharing link flows: A machine learning approach: Fid-Bau Portal

How the built environment affects E-scooter sharing link flows: A machine learning approach

Jin, Scarlett T. / Wang, Lei / Sui, Daniel

Abstract Understanding how the built environment influences e-scooter sharing (ESS) travel behavior is essential for informing effective built environment modifications to promote ESS usage. This study introduces a novel approach by utilizing link flow data, a different type of data compared to previous studies that focused on trip origin or destination information. By incorporating route-based built environment features and employing the gradient boosting decision tree (GBDT) method, we investigate the non-linear relationship between the built environment and ESS link flows. Our findings reveal the five most important variables affecting ESS travel behavior: distance to the city center, types of bike facilities, slope, speed limits, and street trees. Notably, bike facilities with higher levels of physical barriers from vehicle traffic attract more ESS link flows. Additionally, our results affirm the effectiveness of San Francisco's 20 mph commercial corridors in attracting ESS link flows. Our study generates new insights into the influences of the built environment on ESS travel behavior, contributing valuable knowledge that can be used to guide urban design and transportation planning efforts.

Highlights • This study adopts link flow data to measure how the built environment affects ESS travel behavior. • The gradient boosting decision tree (GBDT) method is adopted. • The top five predictors are distance to the city center, bike facilities, slope, speed limits, and street trees. • Bike facilities with higher levels of physical barriers from vehicle traffic attract more ESS link flows. • 20 mph and 30 mph speed limits are the most popular choices, whereas 25 mph is the least favorable one.