|
Abstract
|
OBJECTIVE: The study investigates the relationship between traffic accidents in expressway tunnels and their influencing factors, with the aim of predicting traffic accidents within tunnels and presenting reasonable recommendations to improve tunnel safety.
METHODS: The study utilizes a dataset of 586 traffic accidents occurring exclusively within 8 tunnels along a Guangdong Province expressway from 2017 to 2021. It applies the geometric alignment consistency principle to segment road sections, defines tunnel boundaries based on driving behavior, and employs a Bayesian-modified negative binomial regression model (B-NB model) to identify 6 significant variables from a pool of 17 factors.
RESULTS: The predictive performance of the B-NB model demonstrated similarities to that of the fixed parametric model. This outcome might be attributed to the chosen prior distribution settings and the limited amount of data. Nonetheless, the model effectively captures relationships among variables, leading to improved accuracy in accident prediction and the predictive model achieves a 76.1% accuracy rate.
CONCLUSIONS: Drawing from the estimation results, practical measures are suggested across three dimensions: road geometric alignment design, tunnel traffic safety facilities, and traffic emergency management. These proposals aim to ameliorate the severe consequences of tunnel accidents. Future research will explore an in-depth comparison of estimation results, considering the impact of time and variable correlation on the prediction model by expanding the existing data. This will guide the direction of subsequent research endeavors.
Language: en |