高速公路事故频发,而以往研究未能充分揭示交通流动态特性对事故类型与严重程度的影响。为此研究了基于动态交通流数据的高速公路事故类型与严重程度的预测方法。从高速公路门架数据中提取流量、密度、速度等交通流数据,同时考虑时间特...高速公路事故频发,而以往研究未能充分揭示交通流动态特性对事故类型与严重程度的影响。为此研究了基于动态交通流数据的高速公路事故类型与严重程度的预测方法。从高速公路门架数据中提取流量、密度、速度等交通流数据,同时考虑时间特征以及时间和空间不均匀性特征的数据,与事故数据相匹配构成全样本。建立了基于极端梯度提升树(extrem Gradient Boosting,XGBoost)算法的预测模型,预测事故是否发生、事故类型以及事故严重程度。分别考虑追尾事故和其他事故2种事故类型、有人员伤亡和仅财产损失2种事故严重程度,模型的结果表明:(1)上下游速度差大、低速、路段车流量大且频繁分流、合流条件下交通事故风险较高;(2)低速、路段车辆多且合流、分流交通量大、上下游速度差大的情况下发生追尾事故的风险更高;(3)路段车流量较少且追尾事故发生于周末或夜间可能会增大事故严重程度。将常用机器学习算法与XGBoost算法的预测效果进行对比,XGBoost事故类型预测模型与事故严重程度预测模型的ROC曲线下面积(Area Under Curve,AUC)分别达到了0.76和0.88——相比于序列Logistic、高斯朴素贝叶斯、线性SVM、随机森林以及神经网络等其他常用算法,平均分别提升了0.08和0.24。这表明基于XGBoost建立的模型具有较好的预测性能。研究结果为高速公路路段实时交通流状态预警提供了可靠手段,进而可以提升高速公路行车安全。展开更多
In order to control traffic congestion, many mathematical models have been used for several decades. In this paper, we study diffusion-type traffic flow model based on exponential velocity density relation, which prov...In order to control traffic congestion, many mathematical models have been used for several decades. In this paper, we study diffusion-type traffic flow model based on exponential velocity density relation, which provides a non-linear second-order parabolic partial differential equation. The analytical solution of the diffusion-type traffic flow model is very complicated to approximate the initial density of the Cauchy problem as a function of x from given data and it may cause a huge error. For the complexity of the analytical solution, the numerical solution is performed by implementing an explicit upwind, explicitly centered, and second-order Lax-Wendroff scheme for the numerical solution. From the comparison of relative error among these three schemes, it is observed that Lax-Wendroff scheme gives less error than the explicit upwind and explicit centered difference scheme. The numerical, analytical analysis and comparative result discussion bring out the fact that the Lax-Wendroff scheme with exponential velocity-density relation of diffusion type traffic flow model is suitable for the congested area and shows a better fit in traffic-congested regions.展开更多
文摘高速公路事故频发,而以往研究未能充分揭示交通流动态特性对事故类型与严重程度的影响。为此研究了基于动态交通流数据的高速公路事故类型与严重程度的预测方法。从高速公路门架数据中提取流量、密度、速度等交通流数据,同时考虑时间特征以及时间和空间不均匀性特征的数据,与事故数据相匹配构成全样本。建立了基于极端梯度提升树(extrem Gradient Boosting,XGBoost)算法的预测模型,预测事故是否发生、事故类型以及事故严重程度。分别考虑追尾事故和其他事故2种事故类型、有人员伤亡和仅财产损失2种事故严重程度,模型的结果表明:(1)上下游速度差大、低速、路段车流量大且频繁分流、合流条件下交通事故风险较高;(2)低速、路段车辆多且合流、分流交通量大、上下游速度差大的情况下发生追尾事故的风险更高;(3)路段车流量较少且追尾事故发生于周末或夜间可能会增大事故严重程度。将常用机器学习算法与XGBoost算法的预测效果进行对比,XGBoost事故类型预测模型与事故严重程度预测模型的ROC曲线下面积(Area Under Curve,AUC)分别达到了0.76和0.88——相比于序列Logistic、高斯朴素贝叶斯、线性SVM、随机森林以及神经网络等其他常用算法,平均分别提升了0.08和0.24。这表明基于XGBoost建立的模型具有较好的预测性能。研究结果为高速公路路段实时交通流状态预警提供了可靠手段,进而可以提升高速公路行车安全。
文摘In order to control traffic congestion, many mathematical models have been used for several decades. In this paper, we study diffusion-type traffic flow model based on exponential velocity density relation, which provides a non-linear second-order parabolic partial differential equation. The analytical solution of the diffusion-type traffic flow model is very complicated to approximate the initial density of the Cauchy problem as a function of x from given data and it may cause a huge error. For the complexity of the analytical solution, the numerical solution is performed by implementing an explicit upwind, explicitly centered, and second-order Lax-Wendroff scheme for the numerical solution. From the comparison of relative error among these three schemes, it is observed that Lax-Wendroff scheme gives less error than the explicit upwind and explicit centered difference scheme. The numerical, analytical analysis and comparative result discussion bring out the fact that the Lax-Wendroff scheme with exponential velocity-density relation of diffusion type traffic flow model is suitable for the congested area and shows a better fit in traffic-congested regions.
