The invention proposes a traffic flow prediction method based on a depth map Gaussian process, and the method comprises the following steps: S1, employing an attention kernel function to describe the dynamic dependence between nodes on a topological graph, and employing the dynamic dependence as a covariance function in an aggregation Gaussian process, so as to extract dynamic spatial features; S2, obtaining a time sequence convolution Gaussian process according to the weights at different moments and the convolution function obeying the Gaussian process, and obtaining time sequence features in the traffic data in combination with an aggregation Gaussian process; and S3, constructing a depth map Gaussian process method fusing a Gaussian process and a depth structure according to the aggregation Gaussian process, the time sequence convolution Gaussian process and the Gaussian process with a linear kernel function, and inputting a to-be-predicted data sample into the depth map Gaussian process method to obtain a prediction result. According to the method, complex features are obtained under a small number of samples, traffic flow is predicted, and uncertainty is quantified.

    本发明提出了一种基于深度图高斯过程的交通流量预测方法,包括以下步骤:S1.针对空间依赖中所存在的动态性,采用注意力核函数刻画拓扑图上各结点间的动态依赖,并用作聚合高斯过程中的协方差函数,以提取动态空间特征;S2.由不同时刻下的权重、服从高斯过程的卷积函数得到时序卷积高斯过程,并结合聚合高斯过程,获取交通数据中的时序特征;S3.由聚合高斯过程、时序卷积高斯过程以及具有线性核函数的高斯过程,构建融合高斯过程及深度结构的深度图高斯过程方法,将待预测的数据样本输入至深度图高斯过程方法中,得到的预测结果。该方法实现了在少量样本下获取复杂特征、预测交通流量、量化不确定性。


    Access

    Download


    Export, share and cite



    Title :

    Traffic flow prediction method based on depth map Gaussian process


    Additional title:

    一种基于深度图高斯过程的交通流量预测方法


    Contributors:
    JIANG YUNLIANG (author) / FAN JINBIN (author) / ZHANG XIONGTAO (author) / SHEN QING (author) / LIU YONG (author)

    Publication date :

    2021-11-19


    Type of media :

    Patent


    Type of material :

    Electronic Resource


    Language :

    Chinese


    Classification :

    IPC:    G06F ELECTRIC DIGITAL DATA PROCESSING , Elektrische digitale Datenverarbeitung / 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 / G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS