基于时空交互模型的高速公路季节事故频次影响因素分析

为了深入了解影响高速公路事故频次的显著因素，采集2014年广东省开阳高速公路的事故、道路、交通和气象数据，以曲率和坡度同质性为原则将整条公路划分为154条路段，采用时空交互模型拟合路段季节事故数和道路设计参数、交通特征、气象因素间的内在关系。该模型不仅解释了相邻路段间的空间效应和相邻季节间的时间效应，而且还考虑了时空效应间的相互作用，有助于提高模型的拟合预测性能、减少参数估计偏倚。基于贝叶斯推断的模型估计和评价结果显示：事故数据中存在显著的时空关联和交互效应；时空交互模型比传统层级泊松模型的拟合优度更高；路段长度与事故频次线性相关，而交通量则与事故频次间存在非线性关系；高速公路交通安全性随着中、大型客、货车（三类车）比例的增加而显著提高；路段曲率、坡度越大，交通事故风险越高；风速越高、降水量越多的季节，事故频次将显著上升。研究结果可为高速公路交通安全改善方案的制定提供理论依据。

Seasonal Analysis of Contributing Factors to Freeway Crash Frequency Using a Spatio-temporal Interaction Model

The crash, roadway, traffic, and weather data from Kaiyang Freeway in Guangdong Province in 2014 were collected to comprehensively identify the factors with significant impacts on freeway crash frequency. The whole freeway was split into 154 segments based on homogeneity in horizontal curvature and vertical grade. A spatio-temporal interaction model was developed to approximate the underlying relationship between seasonal crash count on each freeway segment and roadway design attributes, traffic characteristics, and weather-related factors. The proposed model can account for not only the spatial effects across adjacent segments and temporal effects across successive seasons, but also the interaction between the spatial and temporal effects. This is helpful to improve the model's fit and prediction performance and reduce the bias in parameter estimates. The results of the model estimation and assessment via Bayesian inference indicate that: there are significant spatio-temporal correlations and interactions in the crash data; the spatio-temporal interaction model outperforms a traditional hierarchical Poisson model in term of goodness-of-fit; crash frequency is linearly correlated to segment length, but nonlinearly correlated to traffic volume; freeway safety performance would be significantly improved by the increase in percentages of medium/large passenger vehicles and trucks (vehicles in Class 3); crash risk is higher on freeway segments with greater curvatures and higher grades; crash frequency would be significantly increased during seasons with stronger wind and more precipitation. These findings provide theoretical directions for designing countermeasures aimed at improving freeway safety performance.