高速公路路侧事故起数预测模型

发布时间：2018-11-05 12:59

【摘要】：为识别高速公路路侧事故的主要诱导因素,分析路侧事故起数与道路线形和交通条件之间的关系,以广珠(广州-珠海)东线高速公路3年中发生的178起路侧事故为基础,分别采用定长法和不定长法划分路段单元,从道路线形和交通条件中选取11个自变量,采用零堆积负二项回归模型建立路侧事故起数预测模型。选择Vuong检验统计量、对数似然值和AIC信息准则3个指标进行模型的拟合优度检验,选择相对误差和累积残差2个指标进行模型的拟合准确性检验;通过对比分析负二项回归模型和零堆积负二项回归模型的拟合优度和拟合准确性检验结果判断其优劣性,并采用弹性分析确定较优模型中显著自变量对因变量的影响程度。研究结果表明:无论采用定长法划分还是不定长法划分路段单元,零堆积负二项回归模型构建的路侧事故起数预测模型明显优于负二项回归模型;采用零堆积负二项回归模型构建的路侧事故起数预测模型,其定长法划分的路段单元模型的拟合准确性优于不定长法;对于定长法划分的路段单元,基于零堆积负二项回归模型的路侧事故起数预测模型有5个自变量对路侧事故起数均有显著影响,影响程度大小依次为车道数、曲率变化率、曲线比例、曲度和平均纵坡坡度。
[Abstract]:In order to identify the main inducing factors of highway side accidents and analyze the relationship between the number of road side accidents and road alignment and traffic conditions, based on 178 road side accidents on Guangzhu (Guangzhou-Zhuhai) east expressway in 3 years, The method of fixed length and the method of indefinite length are used to divide the road sections, 11 independent variables are selected from the road alignment and traffic conditions, and the prediction model of the initial number of road side accidents is established by using the zero stacking negative binomial regression model. The Vuong test statistic, logarithmic likelihood value and AIC information criterion were selected to test the goodness of fit of the model, and two indexes, relative error and cumulative residual, were selected to test the fitting accuracy of the model. The goodness of fit and the accuracy of fitting of the negative binomial regression model and the zero pile negative binomial regression model are compared and analyzed, and the influence degree of the significant independent variables on the dependent variables in the better model is determined by elastic analysis. The results show that the prediction model of road side accidents based on zero accumulation negative binomial regression model is superior to the negative binomial regression model. The prediction model based on zero-stacking negative binomial regression model has better fitting accuracy than the indefinite length method. For the road sections divided by the method of fixed length, there are five independent variables in the prediction model based on the zero-stacking negative binomial regression model, which have a significant effect on the number of road side accidents, and the degree of influence is the number of lanes and the rate of curvature change. Curve ratio, curvature, and average slope gradient.
【作者单位】：长安大学汽车学院;德邦物流有限公司;云南省交通科学研究院;
【基金】：国家自然科学基金项目(51208052) 高等学校青年骨干教师出国研修项目(201406565054) 中央高校基本科研业务费专项资金项目(310822151024,310822161007)
【分类号】：U491.3

【相似文献】