基于深度学习算法的跨江通道交通流量预测研究

Research on Traffic Flow Prediction of Cross-river Channel Based on Deep Learning Algorithm

交通流量预测是智能交通研究的重要课题,预测结果是交通控制和规划管理的重要依据。基于大规模历史交通流数据与深度学习算法进行跨江通道交通流量预测,首先通过数据降噪、填补缺失值等方式处理数据,然后将数据切分为训练样本和测试样本,基于深度学习算法框架分别利用LSTM和其改进型GRU进行模型训练、优化和效果测试。以南沙大桥和虎门大桥为例,以广东高速公路ETC门架与卡口通行(联网收费)数据为数据源,挖掘跨江通道交通流分布时空特征,对模型进行训练和验证,结果显示相较传统的循环神经网络,LSTM和GRU网络均能够取得较好的训练和预测效果,且在跨江通道交通流量预测中GRU网络具有更好的预测性能。

Traffic flow prediction is one of the key technologies in intelligent transportation systems,and it is also important basis for traffic control and planning management.The cross-river channel traffic flow predication model is constructed based on large-scale historical traffic flow data and deep learning algorithm framework.Firstly,the data is processed by data denoising and generating missing values,and then the data is divided into training samples and test samples.Secondly,the Long Short-Term Memory(LSTM)and its improved Gated Recurrent Unit(GRU)was trained and optimized separately.Lastly,the test sample data was used to test the model.Taking Nansha Bridge and Humen Bridge as channel examples,and Guangdong Expressway Networking Toll Receipt as data source,it analyzed the temporal and spatial characteristics of traffic flow distribution crossing the bridges to train and test the model.The results show that LSTM and GRU can achieve better prediction results and stability compared with the traditional Recurrent Neural Network(RNN).GRU has better performance in the cross-river traffic flow prediction.