Understanding Gender Gap in Bike-Sharing Services via eXplainable Artificial Intelligence

Abstract

Bike-sharing services are emerging as a sustainable urban transportation solution, providing both mobility and environmental benefits. This study explores the impact of various attributes on the gender gap in bike-sharing services in Seoul using the eXplainable artificial intelligence (XAI) technique. Prediction was performed using deep learning and machine learning, and interpretation was conducted using SHapley Additive exPlanations (SHAP). The results showed that the categorical boosting (CatBoost) model achieved the highest performance in predicting gender choice probability, with an F1 score of 0.869, among the five models. Using the CatBoost model, SHAP values were estimated, and two interpretable analyses were conducted: feature analysis and feature dependency analysis. The results indicated that female users prefer bike-sharing services with longer travel times, shorter travel distances, shorter Euclidean distances, smaller commercial areas, higher floating population ratios, and more bicycle-friendly environments. Additionally, bike-sharing stations were classified into those with a high proportion of female users and those with the potential to increase the proportion of female users. These findings provide valuable insights for planners and policy makers in developing policies to create more bicycle-friendly cities.