Identification of accident-prone segments using APIU: A case study on highway safety analysis in China

Abstract

Objective This study proposes the accident point interval unit (APIU) method combined with the characteristics of road traffic accidents. The aim is to automatically identify accident aggregation areas, providing basis for highway design and traffic management. Methods Historical accident data from a secondary highway in Guizhou Province and an expressway in Guangdong Province over 3 to 4 years were analyzed using APIU to identify accident-prone segments. A backpropagation (BP) neural network model was utilized to calculate the weight of the impact of the alignment on the occurrence of the accidents, which were then integrated with evaluation levels to formulate a risk index model. Results The APIU exhibited stability and consistency in identifying accident-prone sections, effectively accounting for the influence of adjacent road sections. The BP neural network model quantified the impact of road alignment on accidents, and the risk index model provided a comprehensive evaluation of road section risk. A significant risk zone was identified within 200 to 300 m before the accident staking point, validating APIU. Conclusions Using the APIU, accident-prone segments can be accurately identified. The risk indexes start rising within a specific range before the accident stakes, suggesting that road accidents are influenced by the geometric alignment preceding the accident point. Based on this insight, highway authorities can implement targeted safety measures and enhance signage in critical areas.