Predicting and interpreting crash severity on rural horizontal curves: An investigation of spatial effects from adjacent curves

Horizontal curves are widely recognized as some of the most crash-prone elements of roadway networks. This study systematically examined how the geometric and spatial characteristics of adjacent curves influence crash severity on rural highway horizontal curves. A three-stage hybrid framework was employed, encompassing the extraction of spatial features, predictive modeling using the XGBoost algorithm, and interpretable analysis of influential factors using the SHAP technique to provide targeted safety countermeasures. To contextualize the modeling approach, the performance of the cost-sensitive XGBoost model was compared with the corresponding LightGBM, Random Forest, and fixed-parameter Binary Logistic Regression implementations on a common fixed holdout subset dataset of the study area. Among these four specific implementations, cost-sensitive XGBoost achieved the highest Balanced Accuracy, Macro-F1, G-Mean, and Macro PR-AUC. In addition to conventional target-curve, environmental, roadway, human, and crash-related factors, the study used SECES to derive potential adjacent-curve predictors. These predictors described the direction and radius of the preceding curve, the ratio of the target-curve radius to the mean radius of the three preceding curves, and curve density within a specified segment. Under the sparse-safe primary specification, travel-lane width, roadway-departure collision, winter season, animal collision, young-driver age, and nighttime were the leading contributors to the model predictions. The ratio of the target-curve radius to the average radius of the three preceding curves also contributed to the fixed-holdout model, although its SHAP pattern was heterogeneous. The adjacent-curve findings therefore require confirmation in larger independent datasets. The resulting framework can assist authorities and engineers in prioritizing safety interventions, evaluating geometric-design concerns, and developing measures to reduce crash severity on rural highway curves.

Authors

Institutions

Publication Details

Journal
Transportation Research Interdisciplinary Perspectives
Published
2026-11-18
DOI
https://doi.org/10.1016/j.trip.2026.102362
Primary Topic
Traffic and Road Safety
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Predicting and interpreting crash severity on rural horizontal curves: An investigation of spatial effects from adjacent curves

Meysam Effati, Mahyar Arabani, Amirhosein Guilanpour
Transportation Research Interdisciplinary Perspectives
Traffic and Road Safety
article

Predicting and interpreting crash severity on rural horizontal curves: An investigation of spatial effects from adjacent curves

Meysam Effati, Mahyar Arabani, Amirhosein Guilanpour
article en

Abstract

Horizontal curves are widely recognized as some of the most crash-prone elements of roadway networks. This study systematically examined how the geometric and spatial characteristics of adjacent curves influence crash severity on rural highway horizontal curves. A three-stage hybrid framework was employed, encompassing the extraction of spatial features, predictive modeling using the XGBoost algorithm, and interpretable analysis of influential factors using the SHAP technique to provide targeted safety countermeasures. To contextualize the modeling approach, the performance of the cost-sensitive XGBoost model was compared with the corresponding LightGBM, Random Forest, and fixed-parameter Binary Logistic Regression implementations on a common fixed holdout subset dataset of the study area. Among these four specific implementations, cost-sensitive XGBoost achieved the highest Balanced Accuracy, Macro-F1, G-Mean, and Macro PR-AUC. In addition to conventional target-curve, environmental, roadway, human, and crash-related factors, the study used SECES to derive potential adjacent-curve predictors. These predictors described the direction and radius of the preceding curve, the ratio of the target-curve radius to the mean radius of the three preceding curves, and curve density within a specified segment. Under the sparse-safe primary specification, travel-lane width, roadway-departure collision, winter season, animal collision, young-driver age, and nighttime were the leading contributors to the model predictions. The ratio of the target-curve radius to the average radius of the three preceding curves also contributed to the fixed-holdout model, although its SHAP pattern was heterogeneous. The adjacent-curve findings therefore require confirmation in larger independent datasets. The resulting framework can assist authorities and engineers in prioritizing safety interventions, evaluating geometric-design concerns, and developing measures to reduce crash severity on rural highway curves.

Transportation Research Interdisciplinary PerspectivesVol. 41
University of Guilan (IR)
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