The invention relates to a traffic flow prediction method based on a graph space-time Transform model considering human mobility. Comprising the following steps: 1) data collection: adopting real traffic data sets in different regions of America, and obtaining data required by an experiment after screening, extracting and preprocessing; 2) spatial feature extraction: designing multi-view graph convolution to reveal different behavior modes of people hidden in the traffic data so as to show an invisible graph structure in the traffic data and fully excavate cross-road network features in space; 3) time feature extraction: performing time feature extraction by using a Transform network with a multi-head attention mechanism; 4) model construction: combining time and space features, and constructing a graph space-time Transform model used for traffic flow prediction; and 5) verifying the data set: performing performance evaluation on the graph space-time Transform model by adopting the traffic data set in the real world, finally determining the effectiveness of the graph space-time Transform model, and applying the graph space-time Transform model to traffic flow prediction. Experiments show that the method has excellent performance for treating similar problems.
基于考虑人类流动性的图时空Transformer模型的交通流预测方法。包括:1)数据收集:本发明采用美国不同地区的真实交通数据集,经过筛选提取预处理以后,得到实验所需的数据;2)空间特征提取:设计多视图图卷积揭示隐藏在交通数据中人们的不同行为模式,从而展现交通数据中看不见的图结构,充分挖掘空间上的跨路网特征;3)时间特征提取:使用带有多头注意力机制的Transformer网络进行时间特征提取;4)模型构建:联合时间和空间特征,构建用于交通流预测的图时空Transformer模型;5)数据集验证:本发明采用真实世界的交通数据集对图时空Transformer模型进行性能评估,最终确定图时空Transformer模型的有效性,并用于交通流预测。以实验表明本发明对于处理类似问题具有出色表现。
Traffic flow prediction method based on graph space-time Transform model considering human mobility
基于考虑人类流动性的图时空Transformer模型的交通流预测方法
2022-04-01
Patent
Electronic Resource
Chinese
Traffic flow prediction method considering space-time similarity
European Patent Office | 2023
|Traffic flow prediction method based on improved space-time Transform
European Patent Office | 2022
|Method for extracting traffic flow prediction model by combining graph Fourier transform TMGCN
European Patent Office | 2023
|European Patent Office | 2022
|Time-space traffic state prediction method based on Transform network
European Patent Office | 2021
|