基于改进卷积网络的高速公路节假日拥堵预测系统

Highway Holiday Jam Prediction System Based on Improved Convolutional Network

针对节假日期间高速公路交通流的特点,构建一种基于改进卷积网络—时空图卷积网络(Spatio-Temporal Graph Convolutional Network,ST-GCN)模型的高速公路节假日拥堵预测系统。该系统基于多种数据算法模型实现节假日拥堵预测,并提供API(Application Program Interface)服务模式进行算法模型调用,使高速公路运营更加智能,管理更加高效。以宁夏回族自治区高速公路为例,将基于ST-GCN模型所得预测结果与采用其他常用模型所得结果相对比,验证该系统的有效性,为高速公路管理和运营相关方提供参考。

Expressway traffic flow prediction is essential for highway planning,operation and management.The highway traffic flow during the holiday period is extremely heavy and traffic jam can happen unexpectedly and last long time.The highway holiday jam prediction system,addressing the complicated situation,is developed to answer the needs of highway management and operation.The ST-GCN(Spatio-Temporal Graph Convolutional Network)is introduced into the prediction model.The model is used to process the highway holiday traffic data from Ningxia Hui Autonomous Region and it is verified through the process.The prediction made by the model is compared with those by other methods to demonstrate its advantage in effectiveness.