Poisson系列交通事故预测模型特点分析及应用  

Analysis of Poisson Traffic Accident Prediction Models and Their Application

作　　者：张铁军[1,2] 唐琤琤[1] 康云霞[3] 赵丹[2] 

机构地区：[1]交通运输部公路科学研究所,北京100088 [2]北京交通大学,北京100044 [3]北京市路政局门头沟公路分局,北京102300

出　　处：《公路交通科技》2010年第6期132-137,共6页Journal of Highway and Transportation Research and Development

基　　金：国家道路交通安全科技行动计划课题(2009BAG13A02)

摘　　要：针对Poisson系列交通事故预测模型中的Poisson、负二项、广义负二项、零堆积Poisson和负二项等多种模型形式及验证方法进行了系统分析和总结,在此基础上结合实际工作经验总结了影响建模结果的重点因素,并进行了实例分析。研究分析表明在实际建模工作中影响建模精度的因素包括模型中变量的选择、模型类型的选择、模型函数形式的确定、因变量类型的选择等。此外,研究人员的建模经验对于最终建立模型的优劣也起很重要的作用。The Poisson traffic accident prediction models,such as Poisson model,negative binomial model,generalized negative binomial model,zero-inflated Poisson and negative binomial model were analyzed,especially the model shapes and validating method.On this basis,the major factors affecting modelling precision were advanced in combination with working experience,and a case study was also performed.The result shows that there are some factors influencing the precision of modelling,such as the selection of variables and model types,the identification of functions,and the selection of dependent variables.The modelling experience of researchers is another important factor influencing the quality of model.

关 键 词：交通工程 事故预测模型 Poisson模型 负二项模型 广义负二项模型 

分 类 号：U491.3[交通运输工程—交通运输规划与管理]