The invention discloses a traffic flow prediction method based on a time-varying fusion graph convolutional network, and the method comprises the steps: 1, data collection and data preprocessing: collecting monitoring record data of multiple sensors, building an original data set, carrying out the data preprocessing work of missing value filling on the original data set, and carrying out the data preprocessing work; dividing the processed data set into traffic flow data of a plurality of periodic modes, and constructing a spatial adjacency matrix according to the connectivity between the paired sensors and the driving distance; and 2, constructing a traffic flow prediction model TFGCN based on a time-varying fusion graph convolutional network, taking the divided traffic flow data of the multiple periodic modes as the input of the model, and transmitting the traffic flow data to the prediction model for training to obtain a traffic flow prediction result. According to the method, the long-term dynamic spatial dependency can be adaptively captured, more comprehensive time-space dependency integration is realized, the method is suitable for predicting various different road traffic flows, and the prediction error of the traffic flows is greatly reduced.
本发明公开了一种基于时变融合图卷积网络的交通流量预测方法,包括:步骤1、数据采集和数据预处理,包括采集多传感器的监测记录数据组建原始数据集,对原始数据集进行缺失值填充的数据预处理工作,将经过处理的数据集划分成多种周期模式的交通流量数据并根据成对传感器之间的连接性和驾车距离构建空间邻接矩阵;步骤2、构建基于时变融合图卷积网络的交通流量预测模型TFGCN,将划分的多种周期模式的交通流量数据作为模型的输入,输送到预测模型中进行训练,得到交通流量的预测结果。该方法能够自适应地捕获长期动态的空间依赖关系,实现更全面的时空依赖关系整合,适用于多种不同的道路交通流量预测并大大降低了交通流量的预测误差。
Traffic flow prediction method based on time-varying fusion graph convolutional network
基于时变融合图卷积网络的交通流量预测方法
2024-06-28
Patent
Electronic Resource
Chinese
European Patent Office | 2023
|Traffic flow prediction method based on adaptive dynamic fusion graph convolutional network
European Patent Office | 2024
|Multibranch Adaptive Fusion Graph Convolutional Network for Traffic Flow Prediction
DOAJ | 2023
|European Patent Office | 2025
|European Patent Office | 2023
|