论文部分内容阅读
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 that the lane change may adversely affect traffic safety, this paper combines the traffic condition data of 64 sections of three expressways in Guangdong Province with the historical data of traffic accidents, and uses the negative binomial distribution forecasting method to establish and calibrate the traffic volume based on the section length , The number of lane changes, the proportion of large car lane change, the number of miles per unit and other five explanatory variables of 10 different combinations of traffic accident prediction model. By calculating the Akaike information criterion for each model, we obtained three optimal models that weighed the model structure (that is, the number of explanatory variables) and the data fitting degree. The results show that although the prediction accuracy of the three optimal prediction models still needs to be improved, the traffic accident prediction model considering the influence of lane change is obviously superior to other models. This shows that the variables related to lane change can be used as effective explanatory variables in traffic accident prediction, and the introduction of this type of variable can better predict the occurrence of highway traffic accidents.