基于时空依赖性的区域路网短时交通流预测模型

Short-term traffic flow forecasting model for regional road network based on spatial-temporal dependency

由于多数交通流预测模型仅利用了目标路段交通流的历史数据,在一定程度上影响了预测效果。为此,该文提出了一种基于时空依赖性的区域路网短时交通流预测模型。首先,根据区域路网各路段间的拓扑关系,将其抽象为明确表征上下游路段关系的树状结构,进而根据上下游通路上交叉口转弯率的多阶分配来量化上下游路段的时空依赖性,并将其用于时空自回归差分移动平均模型(STARIMA)空间权重矩阵的改进,最后利用历史数据对改进后的STA-RIMA模型进行参数标定,并用于短时交通流预测。实验结果表明:经过改进后的STARIMA模型,具有更好的预测效果,为区域路网短时交通流预测提供了一种新的方法。

Most traffic flow forecasting models exploit the historical traffic flow data of the objective link with the forecasting performance being influenced to some extent. This paper presents a short-term regional traffic flow forecasting model based on spatial-temporal dependency. A tree structure was abstracted representing the relations of downstream and upstream links according to the topological relations of regional urban road network. The multiple distribution of turning rates at the intersections on the route from upstream to downstream was used to quantify the spatial temporal dependency with the quantized spatial-temporal dependency then used to modify the spatial weight matrix of STARIMA （space-time autoregressive integrated moving average） model. The parameters of STARIMA model were calibrated using the historical traffic flow data and exploited to the short-term traffic flow forecasting. The experimental results show that the improved STARIMA model can provide better forecasting performance as a new approach for short-term traffic flow forecasting of regional road network.