The invention discloses a traffic flow prediction method based on multi-network feature random fusion of residual connection, which comprises the following steps of: firstly, selecting time period, holidays and festivals and weather information as key factors influencing traffic flow prediction, and taking the time period, the holidays and the weather information as input of a network; meanwhile, due to the fact that gradient disappearance or gradient explosion is prone to occurring in the feature extraction process of a traditional recurrent neural network, an anti-residual connection block in the MobileNet V2 is introduced, and the problem can be effectively solved. And finally, in order to further improve the network characterization capability, providing a random feature fusion strategy, respectively generating corresponding random masks according to the features output by the multi-network model during training, carrying out random fusion on the network output features, outputting the fused features to a prediction network to carry out traffic flow prediction, and outputting a prediction result. According to the invention, effective decision data is provided for implementing more effective traffic management, command scheduling and traffic planning, and the levels of urban road traffic management control, command decision and planning are improved.

    本发明公开了一种基于残差连接的多网络特征随机融合的交通流预测方法,首先,选取时间段、节假日以及天气信息作为影响交通流预测的关键因素,并将这三者作为网络的输入。同时,由于传统的循环神经网络在提取特征时容易出现梯度消失或梯度爆炸等问题,引入MobileNet V2中的反残差连接块,能够有效地缓解上述问题。最后,为进一步提升为网络表征能力,提出了随机特征融合策略,在训练时根据多网络模型输出的特征分别生成对应的随机掩码,对各网络输出特征进行随机融合,将融合特征输出到预测网络中进行交通流预测并输出预测结果。本发明为实施更为有效的交通管理、指挥调度和交通规划提供有效的决策数据,提高城市道路交通管理控制、指挥决策和规划的水平。


    Access

    Download


    Export, share and cite



    Title :

    Traffic flow prediction method of multi-network feature random fusion based on residual connection


    Additional title:

    一种基于残差连接的多网络特征随机融合的交通流预测方法


    Contributors:
    MAO ZHAOYONG (author) / WANG YICHEN (author) / SHEN JUNGE (author) / MENG CHENLIN (author)

    Publication date :

    2023-08-29


    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 multi-feature fusion

    LIU CONG'AI / QIN YUHUA / ZHANG XIAOJING | European Patent Office | 2024

    Free access

    Traffic flow prediction method based on multi-feature extraction and fusion

    TANG YUNXIAO / LI DINGCHENG / LIN ZE | European Patent Office | 2024

    Free access

    Multi-modal traffic flow prediction method based on multi-source data feature fusion

    CHEN LULU / LI SHIJIE / JIANG HUAIGUANG et al. | European Patent Office | 2025

    Free access

    Traffic flow prediction method based on residual network

    ZHANG XU / ZUO CHANGQI / YAN LIANG et al. | European Patent Office | 2023

    Free access

    Traffic flow prediction method of space-time attention graph convolutional network based on multi-feature fusion

    CHEN YAJUN / DING ZHIMING / GUO LIMIN | European Patent Office | 2023

    Free access