The invention discloses a traffic flow prediction method based on an adaptive partial attention diffusion map convolutional neural network, and the method comprises the following steps: preparing a traffic flow data set, dividing the data into three time period segments, i.e., a recent period, a daily period and a weekly period, enabling the data of each period to enter a deep neural network for learning, modeling different traffic modes; before entering the neural network, carrying out standardization processing on the data set; setting two node embedding vectors by using a node embedding algorithm, and constructing a self-adaptive node incidence matrix; defining an attention function between nodes, setting a distance threshold value of the nodes, constructing a node part attention matrix, and fusing the node part attention matrix with the node incidence matrix to obtain a dynamic node incidence matrix; sending the standardized historical data and the dynamic node incidence matrix into a graph convolution layer and a diffusion convolution layer together to obtain processed spatial relationship data; entering a time layer, performing time correlation extraction by using gating diffusion convolution, and obtaining a prediction result through a linear layer; and carrying out weighted fusion on the prediction results of the three periods to obtain a final prediction result. According to the method, the graph neural network is selected, feature capture can be well carried out on graph data, and modeling is carried out on traffic flow data in combination with simplified training of the diffusion convolution layer.

    本发明公开了一种基于自适应部分注意力扩散图卷积神经网络的交通流量预测方法,包括如下步骤:准备交通流量数据集,将数据划分为三个时间周期分段,即最近周期、每日周期和每周周期,每个周期数据分别进入深度神经网络进行学习,建模不同的交通模式;进入神经网络之前,将数据集进行标准化处理;利用节点嵌入算法,设置两个节点嵌入向量,构建自适应的节点关联矩阵;定义节点间的注意力函数,设置节点的距离阈值,构建节点部分注意力矩阵,并将其与节点关联矩阵融合获得动态节点关联矩阵;将标准化的历史数据与动态节点关联矩阵一同送入图卷积和扩散卷积层中,获得处理过空间关系的数据;进入时间层,使用门控扩散卷积进行时间关联提取,通过线性层得到预测结果;将三个周期预测结果进行加权融合得出最终预测结果。本发明选择图神经网络,可以很好的对图数据进行特征捕捉,并结合扩散卷积层简化训练,对交通流量数据进行建模。


    Access

    Download


    Export, share and cite



    Title :

    Traffic flow prediction method based on adaptive partial attention diffusion map convolutional neural network


    Additional title:

    一种基于自适应部分注意力扩散图卷积神经网络的交通流量预测方法


    Contributors:
    ZHANG BOWEN (author) / LI BOHAN (author) / WU JIAQI (author) / XU SHUAI (author) / ZHU DI (author) / LIU YI (author) / HUANG PEIXUAN (author)

    Publication date :

    2024-01-26


    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



    Attention-Based Spatiotemporal Adaptive Graph Diffusion Convolutional Network For Traffic Flow Prediction

    He, Qiansong / Xia, Dawen / Li, Jianjun et al. | Transportation Research Record | 2025


    Traffic flow prediction method based on time attention circulation graph convolutional neural network

    FAN WENDONG / SHU MIN / SONG YUN et al. | European Patent Office | 2023

    Free access

    Traffic flow prediction method based on cyclic attention coupled graph convolutional network

    CHEN LING / CHEN WEIQI | European Patent Office | 2020

    Free access

    Node adaptive learning graph attention neural network traffic flow prediction method

    LOU JUNGANG / HUANG XUXIANG / SHEN QING et al. | European Patent Office | 2024

    Free access

    Traffic flow prediction method based on dynamic adaptive adversarial graph convolutional neural network

    WANG HUI / WANG YU / DU KAI | European Patent Office | 2024

    Free access