融合多因素的“时间齿轮”交通流预测模型

为提高交通流预测的准确度，结合城市道路交通的特点，提出一种融合多因素的“时间齿轮”交通流预测模型（TGM），将模型分为2个模块对交通流的影响因素进行特征挖掘。模块1通过深度分析天气因素的影响，挖掘提取天气因素的特征信息；模块2仿照转轴与齿轮的关系，将目标路段及其邻近交叉口各个方向路段的交通流量数据作为时间轴上的轮齿。模型框架采用正反2个方向都加入注意力机制(attention mechanism)的双向门控循环网络（bidirectional gated recurrent unit，Bi-GRU）。结果表明，TGM模型明显优于多种现有模型；与Bi-GRU模型相比，TGM模型对5，15，25 min的预测精度分别提高了4.75%，6.37%和6.73%。因此，TGM模型能够有效提高交通流预测的准确度，具有更优的中长时预测能力，可为交通组织的优化和交通流理论的研究提供帮助。

"Time gear" traffic flow prediction model with multiple factors integration

In order to improve the accuracy of traffic flow prediction,a time gear traffic flow prediction model (TGM) with multiple factors integration was proposed by combining the characteristics of urban road traffic.The model was divided into two modules for feature mining of traffic flow influencing factors.Module 1 extracted the characteristic information of weather factors by deeply analyzing the influence of weather factors.Module 2 imitated the relationship between rotating shaft and gears.The traffic flow of target roadway and its neighboring intersections were treated as the gear teeth.The model framework adopted the bidirectional gated recurrent unit (Bi-GRU) with attention mechanism in both forward and opposite directions.The results show that the TGM model is significantly better than many existing models.Compared with Bi-GRU model,TGM model improves the accuracy of traffic flow prediction of 5,15 and 25 min by 4.75%,6.37% and 6.73%,respectively.Therefore,TGM model can effectively improve the prediction accuracy of traffic flow,and has better medium and long-term traffic flow prediction ability.It can be helpful for traffic organization optimization and traffic flow theory research.