Citation

Zhang W, Liu T, Yi J. Sustainability (Basel) 2022; 14(18): e11761.

Copyright

(Copyright © 2022, MDPI: Multidisciplinary Digital Publishing Institute)

DOI

10.3390/su141811761

PMID

unavailable

Abstract

Rear-end collisions are caused by drivers misjudging urgent risks while following vehicles ahead in most cases. However, compared with other accident types, rear-end collisions have higher preventability. This study aims to reveal the prone segments and hours of rear-end collisions. First, we extracted 1236 cases from traffic accident records in Harbin from 2015 to 2019. These accidents are classified as property damage accidents, injury accidents and fatal accidents according to the collision severity. Second, density analysis in GIS was used to demonstrate the spatial distribution of rear-end collisions. The collision spots considering the density and severity were visually displayed. We counted the hourly and seasonal distribution characteristics according to the statistical data. Finally, LightGBM and random forest classifier models were used to evaluate the substantial factors affecting accident severity. The results have potential practical value in rear-end collision warning and prevention.

Language: en

Keywords

collision prediction; machine learning model; rear-end collision; spatiotemporal analysis; urban traffic analysis