计算机工程与应用 ›› 2020, Vol. 56 ›› Issue (17): 258-265.DOI: 10.3778/j.issn.1002-8331.1906-0164

• 工程与应用 • 上一篇    下一篇

路网交通流在时空分析背景下的预测研究

李彤伟,王庆荣   

  1. 兰州交通大学 电子与信息工程学院,兰州 730070
  • 出版日期:2020-09-01 发布日期:2020-08-31

Prediction Research on Road Network Traffic Flow in Background of Time and Space Analysis

LI Tongwei, WANG Qingrong   

  1. School of Electronic & Information Engineering, Lanzhou Jiaotong University, Lanzhou 730070, China
  • Online:2020-09-01 Published:2020-08-31

摘要:

深度学习近年来被广泛应用于交通工程领域,针对大型路网中单个路段的交通流预测考虑因素单一、预测精度不高的问题,充分利用长短时记忆(LSTM)网络在时序数据处理方面的优势,结合路网交通流时空分析并运用LSTM模型进行预测。通过对路网中路段检测站点间交通流数据进行相关性计算,并设置不同阈值来选择出代表路段的编号构造原始数据矩阵,对矩阵进行压缩来增加运算效率,最后将压缩矩阵输入模型中进行预测。设置仿真对比实验,验证了提出的方法相较于其他几种模型预测准确率平均可提升11.84%,是一种高效率的交通流预测方法。

关键词: 智能交通系统, 短时交通流预测, 深度学习, 路网, 长短时记忆(LSTM)网络

Abstract:

Deep learning has been widely used in the field of traffic engineering in recent years. For the problem of single traffic flow prediction in a large road network with a single consideration and low prediction accuracy, the advantages of Long Short-Term Memory(LSTM) network in time series data processing are fully utilized. It combines the time and space analysis of road network traffic flow and uses LSTM model for prediction. Correlation calculation is carried out on the traffic flow data between the road segments in the road network, and different thresholds are set to select the original data matrix of the number representing the road segment, and the matrix is compressed to increase the operation efficiency. Finally, the compression matrix is input into the model to predict. The simulation contrast experiment is set up to verify that the proposed method can improve the prediction accuracy by 11.84% compared with other models. It is a highly efficient traffic flow prediction method.

Key words: intelligent transportation system, short-term traffic flow prediction, deep learning, road network, Long Short-Term Memory(LSTM) network