摘要
高速公路交通量的预测是管理部门研究的重要内容,为交通控制和诱导提供数据支撑。针对高速公路交通量的预测问题,引入一种新的基于双向长短期记忆网络(Bidirectional Long Short Time Memory Network,Bi-LSTM)的方法。Bi-LSTM模型将普通的LSTM拆分成为两个方向,前向计算关联历史数据,后向计算关联未来数据,两个方向LSTM不直接连通,将两份数据整合输出作为Bi-LSTM计算单元输出值。实验表明,Bi-LSTM模型相比对比模型预测误差至少优化了4.5%,在非线性交通流数据中具有更好的预测性能和泛化能力。
The forecast of highway traffic volume is an important part of the management department's research,providing data support for traffic control and induction.A new method based on Bidirectional Long Short Time Memory Network(Bi-LSTM)is introduced for the prediction of highway traffic volume.The Bi-LSTM model splits the ordinary LSTM into two directions,the forward calculation associates the historical data,and the backward calculation associates the future data.The two directions LSTM are not directly connected,and the two data integration outputs are output as the Bi-LSTM calculation unit.value.Experiments show that the Bi-LSTM model is optimized by at least 4.5%compared with the prediction model,and has better prediction performance and generalization ability in nonlinear traffic flow data.
作者
温惠英
张东冉
WEN Huiying;ZHANG Dongran(School of Civil and Transportation,South China University of Technology,Guangzhou,Guangdong 510641,China)
出处
《公路工程》
北大核心
2019年第6期51-56,共6页
Highway Engineering
基金
国家自然科学基金资助项目(51378222
51578247)
关键词
交通量预测
高速公路
循环神经网络
双向长短期记忆
traffic volume prediction
expressway
recurrent neural network
two-way long-term and short-term memory
