The invention belongs to the technical field of intelligent traffic, and particularly relates to a traffic flow prediction method based on hybrid deep learning. The invention provides a hybrid deep learning method based on a residual error self-attention (RSA) and a BiGRU-Conv-GRU (RSAB-ConvGRU) network, so as to improve the traffic flow prediction precision. The network is composed of an RSA-Conv GRU module and two bidirectional GRU (Bi-GRU) modules. And the RSA-ConvGRU module comprises a convolution-gating circulation unit (Conv-GRU) module and a residual error self-attention mechanism (RSA) module. Specifically, firstly, Conv-GRU extracts spatial and temporal features using convolution and gating loop units. And secondly, determining contributions of traffic characteristics in different periods by using a residual self-attention (RSA) mechanism, and stabilizing the training process of the network so as to improve the prediction performance of Conv-GRU. And finally, the Bi-GRU module obtains periodic characteristics and forward and backward variance trends in the traffic flow data. Experimental results show that the prediction accuracy of the RSAB-ConvGRU method disclosed by the invention is higher than that of methods such as SVR (Support Vector Register), LSTM (Long Short Term Memory), GRU (Generalized Random Access Unit), DCRNN (Direct Current Recurrent Neural Network), CNN-GRU-Attention, Conv-LSTM, AT-Conv-LSTM, Stacked-LSTM, LSTM-RNN (Long Short Term Memory-Recurrent Neural Network)-Attention and the like.

    本发明属于智能交通技术领域,具体涉及一种基于混合深度学习的交通流量预测方法。本发明提出一种基于残差自注意力(RSA)和BiGRU‑Conv‑GRU(RSAB‑ConvGRU)网络的混合深度学习方法,以提高交通流量预测精度。该网络由RSA‑Conv GRU模块和两个双向GRU(Bi‑GRU)模块组成。RSA‑ConvGRU模块包括卷积‑门控循环单元(Conv‑GRU)模块和残差自注意力机制(RSA)模块。具体来说,首先,Conv‑GRU利用卷积和门控循环单元提取空间和时间特征。其次,利用残差自注意(RSA)机制确定不同时期流量特征的贡献,稳定网络的训练过程,以提高Conv‑GRU的预测性能。最后,Bi‑GRU模块获取交通流量数据中的周期特征和前向、后向方差趋势。实验结果表明,本发明的RSAB‑ConvGRU方法的预测准确性更高于SVR、LSTM、GRU、DCRNN、CNN‑GRU‑Attention、Conv‑LSTM、AT‑Conv‑LSTM、Stacked‑LSTM、LSTM‑RNN withAttention等方法。


    Access

    Download


    Export, share and cite



    Title :

    Traffic flow prediction method based on hybrid deep learning


    Additional title:

    一种基于混合深度学习的交通流量预测方法


    Contributors:
    XIA DAWEN (author) / CHEN YAN (author) / LI HUAQING (author) / LIU HAITAO (author) / ZHANG WENYONG (author) / WANG ZIQIANG (author) / HUO YUJIA (author) / FENG FUJIAN (author) / LU YOUJUN (author) / DENG LI (author)

    Publication date :

    2023-12-12


    Type of media :

    Patent


    Type of material :

    Electronic Resource


    Language :

    Chinese


    Classification :

    IPC:    G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen



    Traffic flow prediction method based on deep learning

    XIE GANG / WANG HAIYING / XIE RUIQI | European Patent Office | 2024

    Free access

    Urban traffic flow prediction method based on deep learning

    WANG RONGXIU / DING YANQIU / ZHENG YAQIAN | European Patent Office | 2025

    Free access

    Lightweight traffic flow prediction method based on deep learning

    ZHENG RUI / DONG CHUNJIAO / GENG QINGQIAO et al. | European Patent Office | 2024

    Free access

    Traffic Flow Prediction Model Based on Deep Learning

    Wang, Bowen / Wang, Jingsheng / Zhang, Zeyou et al. | British Library Conference Proceedings | 2022


    Short-term traffic flow prediction method based on deep learning

    XIAO HONGBO / XIAO JIANHUA / DING LIMING et al. | European Patent Office | 2025

    Free access