The invention provides a traffic prediction method based on dynamic space-time diagram convolution. Traffic prediction is an important component for constructing a smart city, and reasonable traffic prediction can help related departments make important decisions, help people to plan routes and the like. However, due to the fact that the complex spatio-temporal correlation is a challenging task all the time, even if the current research makes progress to a certain extent, the current research still generally pays attention to relation modeling between node pairs and between node historical information, analysis of node properties is neglected, and performance bottleneck is caused. In order to solve the problems, the invention provides a dynamic space-time diagram convolutional neural network (DSTGCN), specifically, a dynamic graph generation module is designed in the invention, and the dynamic graph generation module collects geographical neighbor information and spatial heterogeneity information between node pairs in advance and adaptively fuses the two kinds of information at each time step to generate a new dynamic graph. The dynamic graph module enables the DSTGCN to have the ability to capture dynamic traffic information. Besides, a graph convolution circulation module is constructed, local time dependence is captured on the basis of combining spatial relations, and the local time dependence is used as a supplement of a dynamic graph module to jointly capture the spatial-temporal correlation of the traffic network. According to the method, the effectiveness of the model is verified on the two types of traffic prediction tasks, and the rationality and effectiveness of the DSTGCN are proved through experiments.

    本发明提供了一种基于动态时空图卷积的交通预测方法。交通预测是建设智慧城市的重要组成部分,合理的交通预测可以帮助相关部门作出重要决策、帮助人们出行规划线路等。但由于其复杂的时空相关性一直是一项极具挑战性的任务,即使当前的研究在一定程度上取得了进展,但仍然普遍关注于节点对之间和节点历史信息之间的关系建模,忽略了节点性质的分析,导致了性能瓶颈。为了克服这些问题,本发明提出了一种动态时空图卷积神经网络(DSTGCN),具体来说,在本发明中设计了一个动态图生成模块,它提前采集节点对之间的地理近邻性信息和空间异质性信息,并在每个时间步自适应融合两种信息生成新的动态图。动态图模块使DSTGCN有能力捕捉动态的交通信息。此外,构建了一个图卷积循环模块,在合并空间关系的基础上捕捉局部的时间依赖,它作为动态图模块的补充共同捕捉交通网络的时空相关性。本发明在两类交通预测任务上验证了模型的有效性,实验证明DSTGCN的合理性和有效性。


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    DSTGCN-based traffic prediction method


    Weitere Titelangaben:

    一种基于DSTGCN的交通预测方法


    Beteiligte:
    HU JIA (Autor:in) / LIN XIANGHONG (Autor:in) / WANG CHU (Autor:in) / ZHANG ZHEN (Autor:in) / ZHOU SHASHA (Autor:in)

    Erscheinungsdatum :

    2023-01-06


    Medientyp :

    Patent


    Format :

    Elektronische Ressource


    Sprache :

    Chinesisch


    Klassifikation :

    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 PREDICTION PROGRAM, TRAFFIC PREDICTION APPARATUS, AND TRAFFIC PREDICTION METHOD

    SASAKI HIROSHI / MIURA YOSHIKO / TAKEUCHI SHINICHI et al. | Europäisches Patentamt | 2018

    Freier Zugriff


    TRAFFIC DEMAND PREDICTION DEVICE, TRAFFIC DEMAND PREDICTION METHOD, AND TRAFFIC DEMAND PREDICTION PROGRAM

    TOMIYAMA TOMOE / YANO HIROHITO | Europäisches Patentamt | 2019

    Freier Zugriff

    TRAFFIC PREDICTION SYSTEM, TRAFFIC PREDICTION METHOD, AND PROGRAM

    YAMADA MASAKI / TAMURA YUJI / HORIGUCHI RYOTA et al. | Europäisches Patentamt | 2024

    Freier Zugriff

    TRAFFIC PREDICTION SYSTEM, TRAFFIC PREDICTION METHOD, AND PROGRAM

    YAMADA MASAKI / TAMURA YUJI / HORIGUCHI RYOTA et al. | Europäisches Patentamt | 2024

    Freier Zugriff