The invention relates to the technical field of urban flow prediction, and particularly discloses an urban flow prediction method based on comparative learning. According to the time-space contrast learning pre-training method provided by the invention, through time contrast learning pre-training, the feature distance between the calculated anchor points in the sample set and the positive sample is reduced, and the feature distance between the calculated anchor points in the sample set and the negative sample is increased; the time contrast learning encoder Enctemp obtains the characteristics of the specific traffic distribution condition under the condition that the captured global traffic is similar from the pre-training. According to the method, through spatial contrast learning pre-training, similar region features at the same moment are closer, and a spatial contrast learning encoder Encspace obtains the capability of capturing flow change feature similar regions in a whole city from pre-training. A new sampling algorithm is provided in time contrast learning sample sampling, new positive and negative samples are constructed by using a plurality of positive and negative samples, and the data utilization rate is improved.

    本发明涉及城市流量预测技术领域,特别公开了一种基于对比学习的城市流量预测方法。本发明提出了时空对比学习预训练的方法,通过时间对比学习预训练,计算出来的样本集中的锚点与正样本的特征距离减少,与负样本的特征距离增加,时间对比学习编码器Enctemp从预训练中得到捕捉全局流量相似的情况下具体的流量分布情况的特性。本发明通过空间对比学习预训练,同一时刻的相似区域特征更加接近,空间对比学习编码器Encspatial从预训练中得到捕捉整个城市内流量变化特征相似区域的能力。本发明在时间对比学习样本抽样中提出了新的抽样算法,利用多个正负样本构建新的正负样本,提升了数据利用率。


    Zugriff

    Download


    Exportieren, teilen und zitieren



    Titel :

    City flow prediction method based on comparative learning


    Weitere Titelangaben:

    基于对比学习的城市流量预测方法


    Beteiligte:
    DONG XIANGJUN (Autor:in) / ZHANG XU (Autor:in) / LU WENPENG (Autor:in) / ZHAO LONG (Autor:in)

    Erscheinungsdatum :

    2024-01-02


    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



    Urban traffic flow prediction method and device based on migration comparative learning

    XIAO ZHU / ZHANG TIANRU / JIANG HONGBO et al. | Europäisches Patentamt | 2023

    Freier Zugriff

    Regional flow prediction method based on deep learning

    LI QIQIANG / LIU JIAXIN / SONG WEN | Europäisches Patentamt | 2021

    Freier Zugriff

    Traffic flow prediction method based on deep learning

    ZHANG SHICHAO / DUAN HONGXIANG / ZHANG CHENGYUAN | Europäisches Patentamt | 2022

    Freier Zugriff

    Traffic flow prediction method based on hybrid deep learning

    XIA DAWEN / CHEN YAN / LI HUAQING et al. | Europäisches Patentamt | 2023

    Freier Zugriff

    Short-term traffic flow prediction method based on deep learning

    LI ZHUANGZHUANG / GUI ZHIMING / GUO LIMIN et al. | Europäisches Patentamt | 2020

    Freier Zugriff