The invention relates to a traffic prediction method using a dynamic multi-level convolutional network, and the method comprises the steps: constructing a function similarity-based graph structure to obtain a function similarity graph, and obtaining a spatial distance-based graph structure to obtain a spatial distance graph; through a self-attention mechanism, according to the input features of each convolution sub-network, dynamically adjusting different types of spatial relationship weights so as to carry out spatial convolution processing, and obtaining a first feature map; historical time in the first feature map is divided into time relations of four dimensions of a long term, a middle term, a short term and a global, and fusion convolution processing is carried out on the time relations of the four dimensions to obtain a second feature map; encoding the space-time position information in the second feature map, and performing attention guidance on the encoded space-time information to obtain space-time learning features; and splicing the space-time learning features output by the at least one convolutional sub-network to obtain a traffic prediction value at a future target moment. The method improves the accuracy of traffic prediction.

    本发明涉及一种利用动态多层次卷积网络的交通预测方法,包括:构建基于功能相似性的图结构得到功能相似性图,并获取基于空间距离的图结构得到空间距离图;通过自注意力机制,根据每个卷积子网络的输入特征动态地调整不同类型的空间关系权重以进行空间卷积处理,得到第一特征图;将第一特征图中的历史时间划分为长期、中期、短期和全局四种维度的时间关系,并对四种维度的时间关系进行融合卷积处理,得到第二特征图;对第二特征图中时空位置信息进行编码,并对编码后的时空信息进行注意力引导,得到时空学习特征;将至少一个卷积子网络输出的时空学习特征进行拼接,得到未来目标时刻的交通预测值。该方法提高了交通预测的准确性。


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    Traffic prediction method using dynamic multi-level convolutional network


    Weitere Titelangaben:

    一种利用动态多层次卷积网络的交通预测方法


    Beteiligte:
    LI HE (Autor:in) / JIN DUO (Autor:in) / LI XUEJIAO (Autor:in) / CHENG PENGFEI (Autor:in) / HUANG JIANBIN (Autor:in)

    Erscheinungsdatum :

    2024-01-26


    Medientyp :

    Patent


    Format :

    Elektronische Ressource


    Sprache :

    Chinesisch


    Klassifikation :

    IPC:    G06Q Datenverarbeitungssysteme oder -verfahren, besonders angepasst an verwaltungstechnische, geschäftliche, finanzielle oder betriebswirtschaftliche Zwecke, sowie an geschäftsbezogene Überwachungs- oder Voraussagezwecke , DATA PROCESSING SYSTEMS OR METHODS, SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL, SUPERVISORY OR FORECASTING PURPOSES / G06N COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS , Rechnersysteme, basierend auf spezifischen Rechenmodellen / G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS



    Traffic flow prediction method based on multi-view dynamic graph convolutional network

    HUANG XIAOHUI / YE YUMING / LING JIAHAO et al. | Europäisches Patentamt | 2022

    Freier Zugriff

    Dynamic multi-graph convolutional network traffic flow prediction method based on graph transformation

    HU YONGLI / PENG TING / GUO KAN et al. | Europäisches Patentamt | 2022

    Freier Zugriff


    Temporal Multi-Graph Convolutional Network for Traffic Flow Prediction

    Lv, Mingqi / Hong, Zhaoxiong / Chen, Ling et al. | IEEE | 2021


    Traffic Demand Prediction Based on Dynamic Transition Convolutional Neural Network

    Du, Bowen / Hu, Xiao / Sun, Leilei et al. | IEEE | 2021