Cluster-aware explainable AI for motorcycle accident hotspot prediction: a spatial modeling framework

This study develops an interpretable, spatially-informed machine learning framework to identify motorcycle accident hotspots in Suwon and Seongnam, South Korea. By integrating XGBoost-based probability modeling, spatial clustering using Local Indicators of Spatial Association (LISA), and SHAP (SHapley Additive exPlanations) for feature attribution, the framework captures both the magnitude and spatial structure of accident risk. Grid-based classification models achieved high predictive performance, with government-designated hotspot locations receiving predicted probabilities exceeding 0.8. Cluster-specific SHAP analysis revealed that high-risk zones were driven by road density, pedestrian infrastructure, and elevation complexity, whereas low-risk zones were characterized by natural land cover and sparse built environments. These findings demonstrate the utility of combining spatial statistics and interpretable AI to inform zone-specific traffic safety interventions. The proposed framework is readily transferable to other cities and traffic modes, offering a scalable approach to data-driven risk management.

Authors

Institutions

Publication Details

Journal
Transportation Letters
Published
2026-11-17
DOI
https://doi.org/10.1080/19427867.2026.2756488
Primary Topic
Traffic and Road Safety
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Cluster-aware explainable AI for motorcycle accident hotspot prediction: a spatial modeling framework

박진혁, Chulmin Jun, Dongbeom Kim, Yuhan Han et al.
5 links
Transportation Letters
Traffic and Road Safety
article

Cluster-aware explainable AI for motorcycle accident hotspot prediction: a spatial modeling framework

박진혁, Chulmin Jun, Dongbeom Kim, Yuhan Han, Suyun Lee, Hyemin Kim
article en

Abstract

This study develops an interpretable, spatially-informed machine learning framework to identify motorcycle accident hotspots in Suwon and Seongnam, South Korea. By integrating XGBoost-based probability modeling, spatial clustering using Local Indicators of Spatial Association (LISA), and SHAP (SHapley Additive exPlanations) for feature attribution, the framework captures both the magnitude and spatial structure of accident risk. Grid-based classification models achieved high predictive performance, with government-designated hotspot locations receiving predicted probabilities exceeding 0.8. Cluster-specific SHAP analysis revealed that high-risk zones were driven by road density, pedestrian infrastructure, and elevation complexity, whereas low-risk zones were characterized by natural land cover and sparse built environments. These findings demonstrate the utility of combining spatial statistics and interpretable AI to inform zone-specific traffic safety interventions. The proposed framework is readily transferable to other cities and traffic modes, offering a scalable approach to data-driven risk management.

Transportation Letters
University of Seoul (KR)
University of Seoul
Sustainable cities and communities, Industry, innovation and infrastructure
Traffic and Road Safety
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon