Analyzing rear-end crash severity for a mountainous expressway in China via a classification and regression tree with random forest approach

Abstract

To understand the contributory factors to rear-end accident severity on mountainous expressways, a total of 1039 rear-end accidents, occurring on G5 Jingkun Expressway from Hechizhai to Qipanguan in Shaanxi, China over the period of 2012 to 2017, were collected, and a non-parametric Classification and Regression Tree (CART) model was used to explore the relationship between severity outcomes and driver factors, vehicle characteristics, roadway geometry and environmental conditions. Then the random forest model was introduced to examine the accuracy of variable selection and rank their importance. The results show that driver’s risky driving behaviours, vehicle type, radius of curve, angle of deflection, type of vertical curve, time, season, and weather are significantly associated with rear-end accident severity. Speeding and driving while drunk and fatigued are more prone to result in severe consequences for such accidents and driving while fatigued is found to have the highest fatality probability, especially during the night period (18:00–24:00). The involvement of heavy trucks increases the injury probability significantly, but decreases the fatality probability. In addition, adverse weather and sharp curve with radius less than 1000mare the most risk combination of factors. These findings can help agencies more effectively establish stricter regulations, adopt technical measures and strengthen safety education to ensure driver’s driving safety on mountainous expressways for today and tomorrow.

Keywords

• rear-end accidents
• mountainous expressway
• accident severity
• cart model
• random forest