Kim, Hyeonseo (Korea Transport Institute), Hong, Shngjin (The Korea Transport Institute), sehyun, tak (korea transport institute), Seong, MinGyo (KOREA TRANSPORT INSTITUTE)

Spatial Distribution and Influential Factors of Demand-Responsive Mobility Services Using SHAP-Based Machine Learning

Scheduled for presentation during the Regular Session "S34b-Data-Driven Optimization and Governance in Intelligent Urban Mobility" (FR-EA-T34), Friday, November 21, 2025, 14:50−14:50, Surfers Paradise 1

2025 IEEE 28th International Conference on Intelligent Transportation Systems (ITSC), November 18-21, 2025, Gold Coast, Australia

This information is tentative and subject to change. Compiled on October 18, 2025

Abstract

Amid growing concerns about the sustainability of public transportation in aging and depopulating regions, particularly in small towns and rural areas, traditional fixed-route bus services face increasing operational deficits and declining usage. In response, demand-responsive transport (DRT) models are being considered as flexible and cost-effective alternatives. However, local governments often struggle with the fundamental planning question of determining which areas should transition from fixed-route buses to DRT services. This study proposes a nationwide, interpretable machine learning framework to support such decisions by classifying administrative units into appropriate public mobility service types—fixed-route bus, DRT-bus, or DRT-taxi—based on nationally available demographic, infrastructural, and accessibility indicators. The dependent variable is defined using actual service usage data, while input variables include population structure, car and driver license ownership, and walking distances to key facilities. Models were trained using LightGBM, CatBoost, and XGBoost, with LightGBM selected based on 10-fold cross-validation performance. SHAP analysis was used to interpret feature importance, and Partial Dependence Plots (PDP) revealed threshold effects that guide service transition decisions. The results highlight population density, elderly ratio, and accessibility to public infrastructure as key predictors of suitability. This study provides a scalable, data-driven framework to support local transit planning and offers actionable, interpretable guidelines for replacing unsustainable bus routes with demand-responsive services. It contributes to both academic understanding and practical policy design by demonstrating how explainable AI can be used to enhanc