高速公路大区段交通事故预测模型研究

发布时间：2019-02-16 09:10

【摘要】：针对交通事故记录位置精度较低和路段划分过短可能引起的分析结论偏差问题,探讨了大区段划分条件下高速公路交通事故预测模型的构建方法。结合辽宁省5条高速公路的交通事故数据和道路交通条件数据,提出了基于高速公路天然节点(互通式立交和服务区)的大区段划分方法。构建了处理大区段内指标异质性的模型变量,应用积分-微分方法确定了变量模块,进而利用负二项回归方法建立了高速公路交通事故预测模型。研究表明:在大区段条件下,年平均日交通量、累积纵坡和标志密度等变量的事故预测模块为Hoerl函数,累积曲率为幂函数,而挖方段比例为指数函数;路段长度、年平均日交通量、累积纵坡和标志密度等因素对大区段上的交通事故发生具有显著影响,且考虑这些变量的模块式模型具有较高的预测精度。
[Abstract]:In order to solve the problem of the deviation of analysis conclusion caused by the low accuracy of traffic accident record and the short division of road section, this paper discusses the method of constructing the traffic accident prediction model of expressway under the condition of large section division. Based on the traffic accident data and road traffic condition data of 5 highways in Liaoning Province, a method of dividing large sections based on natural highway nodes (interchange and service area) is proposed. The model variables which deal with the heterogeneity of indexes in large sections are constructed. The variable module is determined by using the integral-differential method, and then the traffic accident prediction model of expressway is established by using the negative binomial regression method. The results show that under the condition of large area, the accident prediction module of the annual average daily traffic volume, cumulative longitudinal slope and sign density is Hoerl function, the cumulative curvature is power function, and the proportion of excavated section is exponential function. The road length, annual average daily traffic volume, cumulative longitudinal slope and sign density have a significant influence on the occurrence of traffic accidents in large sections, and the modular model with these variables has a high prediction accuracy.
【作者单位】：哈尔滨工业大学交通科学与工程学院;
【基金】：中国博士后科学基金面上项目(2016M590285) 黑龙江省博士后基金项目(LBH-Z15092)
【分类号】：U491.31

【相似文献】