基于局部路网空间结构特征的无检测器路段交通流预测方法

城市路网中存在大量尚未布设交通检测器的路段，其交通流数据难以获取，不利于开展精准路网管理，为此提出了利用局部路网空间结构特征预测无检测器路段交通流量的方法。基于有检测器路段的海量交通流数据，分析局部路网空间结构特征与路段交通流量之间的相关性；根据路网拓扑关系使用多元线性回归算法估计所有的有检测器交叉口交通流分配权重，并使用多元线性回归算法进一步挖掘局部路网空间结构特征对交通流分配权重的影响；结合空间特征影响度系数、无检测器路段所在的局部路网的空间结构特征及相邻路段的交通流，对无检测器路段进行交通流预测。实验结果表明，路段道路类型、相邻路段数量及相邻路段道路类型这3类局部路网空间结构特征与路段交通流量相关性显著，采用基于空间特征影响度系数对局部路网中只有单个相邻上游和具有多个相邻上游的无检测器路段进行预测，发现其平均误差分别在8%和22%左右。 

A Method of Traffic Flow Prediction for Road Segments without Detectors Based on Spatial Structure of Local Network

Most links in an urban road network are not monitored by any traffic detector. Lack of traffic flow data has seriously hindered the performance of traffic management programs. In this regard, this paper proposes a traffic flow prediction method for road segments without a detector(RSWD)based on the spatial structure of the local road network. The correlation between the spatial structure of the local network and the traffic flow of the links is analyzed based on the big data of traffic flow. According to the topology of the local road network, multiple linear regression is used to estimate traffic flow assignment weights using data from links with detectors, and to analyze the impacts of the spatial structure of the local road network on traffic flow assignment weights. Then, a method for estimating the traffic flow of road segments without any detector is proposed by considering the spatial structure of the local network and traffic flow of adjacent links. The results show that a significant correlation is found among links of traffic flow and its functional class, and the number and functional class of its adjacent links. The average error of traffic flow prediction based on the proposed model is about 8% and 22% for the RSWD connected with one and several adjacent upstream links in the local road network, respectively.