The invention discloses a traffic flow prediction method based on a de-noising attention-enhanced cyclic multi-graph convolutional network, and the method comprises the following steps: collecting the original traffic flow data of a road network to obtain a noisy flow data set, and carrying out the preprocessing operation to obtain a noiseless flow data set; aligning the time scales of the two data sets and splitting the data sets into a training set and a test set according to a fixed proportion; constructing a de-noising attention enhancement cyclic multi-graph convolution network, wherein the network structure is composed of a de-noising diffusion gating attention fusion module and a cyclic multi-graph convolution module; and training and testing the de-noising attention-enhanced cyclic multi-graph convolutional network by using noisy and noiseless flow data sets, and using the de-noising attention-enhanced cyclic multi-graph convolutional network for traffic flow prediction. According to the method, a denoising probability diffusion model and a multi-head attention mechanism are introduced to enhance lost potential trend features in a data preprocessing process, and a cyclic multi-graph convolution module is utilized to realize extraction of traffic flow dynamic multi-graph spatial-temporal features, so that the traffic flow prediction precision is further improved.
本发明公开一种基于去噪注意力增强循环多图卷积网络的交通流预测方法,包括以下步骤:采集路网原始交通流量数据得到有噪声流量数据集,经过预处理操作得到无噪声流量数据集,将两种数据集时间尺度对齐并按照固定比例拆分为训练集和测试集;构造去噪注意力增强循环多图卷积网络,网络结构由去噪扩散门控注意力融合模块、循环多图卷积模块组成;利用有噪声和无噪声流量数据集训练和测试去噪注意力增强循环多图卷积网络,并用于交通流预测。本发明通过引入去噪概率扩散模型和多头注意力机制来增强数据预处理过程中丢失的潜在趋势特征,利用循环多图卷积模块实现交通流动态多图时空特征的提取,实现交通流预测精度的进一步提高。
Traffic flow prediction method based on de-noising attention enhancement cyclic multi-graph convolutional network
基于去噪注意力增强循环多图卷积网络的交通流预测方法
2024-08-16
Patent
Electronic Resource
Chinese
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 cyclic attention coupled graph convolutional network
European Patent Office | 2020
|European Patent Office | 2024
|European Patent Office | 2023
|European Patent Office | 2023
|Transportation Research Record | 2025
|