基于机器学习的城市快速路速度-密度关系模型

Speed-density relationship model of urban expressway based on machine learning

为进一步提高速度-密度关系模型的精度以及更加精确刻画当前道路交通流动态变化特性,采用基于机器学习的方法对北京市三环路实测数据进行分析。首先对道路实测数据进行预处理,再采用层次聚类法确定不同交通流相位临界密度,最后利用传统速度-密度模型、BP神经网络以及局部加权回归进行对比分析并建立分相位的道路交通流速度-密度关系模型。利用现场实测数据对模型进行测试,结果表明:采用层次聚类可以为每个相位计算出更准确的分界点;与传统模型相比较,采用基于机器学习的方法为每个相位估计速度-密度关系模型能够获得更高的精度;对道路交通流进行分相位,建立速度-密度关系模型比基本图论的方法更能准确捕捉动态交通流的变化趋势。

In order to further improve the accuracy of the speed-density relationship model and more accurately characterize the dynamic characteristics of current road traffic flow,this paper used a method based on machine learning to analyze the measured data of the Third Ring Road in Beijing.This paper first preprocessed the road actual measured data;then used the hierarchical clustering method to determine the critical density of different traffic flow phases;Finally,the traditional speed-density model,BP neural network and local weighted regression are used for comparison and analysis,and a phased road traffic flow speed-density relationship model is established.The model was tested with the field data,and the results showed that the hierarchical clustering method can calculate more accurate dividing points for each phase;compared with the traditional model,the method based on machine learning can estimate the speed density relationship model for each phase with higher accuracy;the speed density relationship model is established by dividing the phases of the road traffic flow,which is more accurate than the basic diagram in capturing the trend of dynamic traffic flow.