The invention discloses a dynamic space-time neural network traffic flow prediction method based on an attention mechanism. The method comprises the following steps: modeling a flow input component, a flow output component and a period component; constructing a local dynamic prediction module; constructing a global correlation prediction module; and constructing a fusion prediction module. Firstly, modeling is carried out on recent cycle dependence, daily cycle dependence and weekly cycle dependence of traffic data, and urban traffic high-dimensional features are extracted on each component by using a three-dimensional convolutional neural network. Then, an improved residual structure is used for capturing the correlation degree of a long-distance area to a prediction area, and a space attention mechanism and a time attention mechanism are fused to capture the dynamic correlation between traffic flows in different areas and different time periods; and finally, performing weighted fusion on the output of the three components by using a method based on a parameter matrix to obtain a prediction result. Experimental analysis and results show that compared with an existing prediction method, the method has better prediction precision and robustness.

    本发明公开了一种基于注意力机制的动态时空神经网络交通流量预测方法,步骤为:对流输入、流输出与周期组件进行建模;构建局部动态预测模块;构建全局相关性预测模块;构建融合预测模块。首先,通过对交通数据的最近周期依赖、日周期依赖和周周期依赖进行建模,在每个分量上使用三维卷积神经网络提取城市交通高维特征。然后,使用改进的残差结构捕捉远距离区域对预测区域的相关度,融合空间注意力和时间注意力机制捕捉不同区域不同时间段上的交通流量之间的动态相关性。最后,使用基于参数矩阵的方法对三个分量的输出进行加权融合,得到预测结果。实验分析和结果表明,与现有预测方法相比,本发明具有更好的预测精度和鲁棒性。


    Access

    Download


    Export, share and cite



    Title :

    Dynamic space-time neural network traffic flow prediction method based on attention mechanism


    Additional title:

    一种基于注意力机制的动态时空神经网络交通流量预测方法


    Contributors:
    MENG XIANGFU (author) / XU RUIHANG (author) / FAN HONGYU (author) / MA RONGGUO (author)

    Publication date :

    2023-01-31


    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 / G06F ELECTRIC DIGITAL DATA PROCESSING , Elektrische digitale Datenverarbeitung / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen



    Traffic flow prediction method based on time-space gated attention map neural network

    XU JIE / GENG ZILI / WU RONGSEN et al. | European Patent Office | 2024

    Free access

    Traffic flow prediction method based on space-time attention mechanism

    WANG JIAYING / YANG HENG / SHAN JING et al. | European Patent Office | 2024

    Free access

    Traffic flow prediction method based on space-time attention network

    MA CHUANG / YAN LI / LIU SHUAIWU et al. | European Patent Office | 2023

    Free access

    Traffic flow prediction method and device based on space-time attention graph neural network

    HUANG WENBING / YUAN JIRUI / TIAN CHUJIE et al. | European Patent Office | 2024

    Free access

    Traffic flow prediction method based on cyclic space-time attention mechanism

    XU XIAO / ZHANG LEI / LIU BAILONG et al. | European Patent Office | 2024

    Free access