高速公路交通事故起数时空分析模型

Temporal-spatial analysis model of traffic accident frequency on expressway

为了分析交通事故起数与时间、道路空间结构及交通运行环境等潜在影响因素之间的关系,从时间和空间角度选择9个自变量,分别从路段长度一致和路段坡度一致2个角度,构建交通事故起数时段、周日和月分布模型。以某典型交通事故多发段为例,分别运用泊松回归模型、负二项回归模型、零堆积泊松回归模型和零堆积负二项回归模型拟合交通事故起数时段、周日和月分布模型,根据模型的拟合优度检验,分别确定3个模型的最佳形式,从而构建交通事故起数时空分析模型。研究结果表明:从AIC准则和BIC准则来看,基于路段长度一致的交通事故起数时段、月分布模型采用负二项回归模型拟合效果较好,其他模型选择泊松回归模型拟合效果较好;基于路段长度一致的交通事故起数时段、周日、月分布模型的预测误差小于基于路段坡度一致的交通事故起数时段、周日、月分布模型。

In order to analyze the relationships among traffic accident frequency and potential influencing factors such as time,road space structure and traffic running environment,nine independent variables were selected from the aspects of time and space,two kinds of section divided methods were adopted,which were fixed-length consistent segment and longitudinal grade consistent segment,and the hourly,weekly and monthly distribution models of traffic accident frequency were constructed.A typical accident-prone section was selected,and Poisson regression model,negative binomial regression model,zero-inflated Poisson regression model and zero-inflated negative binomial regression model were used to fit hourly,weekly and monthly distribution models respectively.The best forms of three models were determined,and the temporal-spatial analysis model of traffic accident frequency was established based on the goodness of fit test.Analysis result shows that the fitting effect of negative binomial regression model is better for traffic accident hourly and monthly distribution models based on fixed-length consistent segment from the views of AIC and BIC,and the fitting effect of Poisson regression model is better for other models.The prediction errors of traffic accident hourly,weekly and monthly distribution model based on fixed-length consistent segment are less than those of longitudinal grade consistent segment.4 tabs,15 refs.