Interpretable machine learning uncovers urban morphology–multi-flow coupling in metro hubs: a lightGBM-SHAP approach
Abstract
Urban hubs underpin sustainable city functioning, yet how urban morphology shapes multiple flows remains unclear. Using interpretable machine learning, this study analyzed 1,245 metro-station hubs in Shanghai, Guangzhou and Shenzhen. A Light Gradient Boosting Machine (LightGBM) model integrated with SHapley Additive exPlanations (SHAP) quantified the nonlinear and threshold effects of seven morphological indicators on four flows—human, logistics, information and financial. Optimal performance emerges with a floor-area ratio 2.5–3.5, building-coverage ratio of 25%–45%, and mean building height ≈ 30 m. Mean building volume shows a bimodal effect: both compact (
Fields of Study & Subjects
Urban Design and Spatial Analysis
Human Mobility and Location-Based Analysis
Urban Transport and Accessibility
Publication details
- Journal / source
- Transportmetrica B Transport Dynamics
- Publisher
- Informa UK Limited
- Volume
- 14
- Issue
- 1
- Publication date
- 3 January 2026
- Language
- EN