基于组合模型的短时交通流预测方法

Short-Term Traffic Flow Forecasting Method Based on Combination Model

准确、实时的交通流预测对交通规划、交通管理和交通控制具有重要意义。然而,由于道路网络拓扑结构约束和交通流随时间动态变化的空时相关特性,交通流预测仍然具有挑战性。为了同时捕获交通流的空间和时间相关性,提出一种将图卷积网络(GCN)和门控循环单元(GRU)相结合的组合模型方法。利用GCU能够灵活处理图结构数据的优点来捕捉各个路段的空间特征,继而发挥GRU在处理时间序列方面的优势挖掘交通流的内在时间规律,空时融合后得到最终预测结果。利用美国交通研究数据实验室的高速公路交通数据对该模型进行仿真验证,结果表明,所提出的GCN-GRU组合模型方法具有更高的预测精度,预测结果优于自回归积分滑动平均(ARIMA)模型和GRU模型等基准预测方法。

Accurate and real-time traffic flow forecasting is of great significance to traffic planning,traffic management and traffic control.However,traffic flow forecasting remains challenging owing to the constraints of the road network topology and the spatial-temporal correlation of traffic flow with time.To capture the spatial and temporal dependence of traffic flow simultaneously,a novel neural network-based traffic forecasting method combining graph convolutional network(GCN)and gated recurrent unit(GRU)is proposed.The advantage of GCN that can deal with graph structure data flexibly was used to capture the spatial characteristics of each road section.The advantage of gated cycle unit in dealing with time series was used to mine the internal time laws of traffic flow.The final prediction result was obtained after spatial-temporal fusion.The model was simulated and verified by highway traffic data from the American Transportation Research Data Laboratory.The results show that the proposed GCN-GRU combined model method has higher prediction accuracy,and the prediction results are better than the benchmark prediction methods such as autoregressive integrated moving average(ARIMA)model and the GRU model.