Abstract: Crash frequency and the influencing factors relating to freeways are changing over time, which means that crash prediction models developed in the past may not be suitable for current traffic conditions. In order to make sure that the implemented safety models fit the current traffic condition, this study proposed a comparative analysis on the basis of freeway datasets in 2017 and 2020 collected from Suzhou, China. Considering the spatial correlation among analysis units and the hierarchical data structure involved, a Bayesian conditional autoregressive negative binomial (CAR-NB) model and a Bayesian hierarchical CAR-NB model were used to explore the varied effects on safety of various influencing factors. The results showed that 1) the HCAR-NB model outperformed the CAR-NB model in prediction accuracy and 2) the number of crashes was significantly correlated with the average speed, speed variance, segment length, and several geometric design features. In addition, Bayesian inference with informative priors was used to update the HCAR-NB model to improve its goodness-fit and efficiency. Based on the modeling results, the potential for safety improvement (PSI) method was used to identify hotspots for the two years. The results confirmed that the hotspots spatiotemporally shift among the freeways. The proposed crash prediction model and model updating method are expected to help local traffic police develop a better understanding of the changes in contributing factors and therefore make informed decisions about safety countermeasures.
Xuesong Wang*, Qi Zhang, Xiaohan Yang, Yingying Pei, Jinghui Yuan, Chao Wang, Juntao Wang. Traffic Safety Analysis and Model Updating for Freeway Using Bayesian Method. Transportation Research Board 101th Annual Meeting, Washington D.C., USA, 2022. 1.9-13.