考虑车道变换影响的高速公路交通事故预测模型研究  

Research on Accident Prediction Models for Freeways Considering Effect of Lane Change Maneuvers

作　　者：孟祥海[1] 郑来[1] 史永义[1] 

机构地区：[1]哈尔滨工业大学交通科学与工程学院,黑龙江哈尔滨150090

出　　处：《公路交通科技》2014年第6期121-126,共6页Journal of Highway and Transportation Research and Development

基　　金：广东省交通运输厅科技项目(2012-2013)

摘　　要：考虑车道变换可能对交通安全造成不利影响,结合广东省3条高速公路64个路段的交通运行状况数据和交通事故历史数据,利用负二项分布预测方法,建立并标定了基于交通量、路段长度、车道变换次数、大型车变道比例、单位里程变道次数等5个解释变量10组不同组合的交通事故预测模型。通过计算各组模型的Akaike信息量准则指标,得到了3组权衡了模型结构(即解释变量数量)和数据拟合度的最优模型。结果表明,虽然3组最优预测模型的预测精度仍有待提高,但是考虑车道变换影响的交通事故预测模型明显优于其他模型。这说明与车道变换相关的变量可以作为交通事故预测的有效解释变量,并且引入该类型变量可以更好地预测高速公路交通事故的发生。Considering the possible adverse effects of lane change maneuvers on traffic safety, based on the traffic operation condition data and traffic accident data collected from 64 segments of 3 freeways in Guangdong Province, 10 accident prediction models are established and calibrated by employing the negative binomial distribution prediction method. Different combinations of traffic volume, length of segment, number of lane change maneuvers, percentage of lane change maneuvers of oversized vehicles, and number of lane change maneuvers per kilometer are taken as the explanatory variables in these models. Three best models, which balanced the model structure （i. e. , the number of explanatory variables） and the data fitting degree, are obtained through calculating the Akaike＇s information criterion of the models. The results show that although the overall prediction accuracy of the 3 best models still needs to be improved, the traffic accident prediction models considering the effect of lane change maneuvers are obviously better than other models. This implies that the variables related to lane change maneuvers can be taken as the valid explanatory variables in predicting freeway accidents, and the introduction of this type of variables could better predict the occurrence of accidents in expressway.

关 键 词：交通工程 事故预测模型 负二项分布 车道变换 Akaike信息量准则 

分 类 号：U491.31[交通运输工程—交通运输规划与管理]