The invention discloses a complex traffic network flow prediction method based on a deep learning model, and the method comprises the steps: building a dynamic graph for the historical traffic data according to the historical traffic data of a road network and the clustering data after spectral clustering; embedding a historical sequence into a potential dimension through a convolution model, and performing static fusion on two different levels of features; a STODE-Block is adopted for the features of the two levels to capture the dependency relationship between time and space, continuous unified modeling is conducted on the time and the space, and feature representation after learning is obtained; performing second dynamic fusion on the feature representations of the regions and the nodes through a dynamic fusion module to obtain final feature output; and combining the two fusion results by adopting a jump connection mode, and then outputting a final prediction result through a ReLU activation function. According to the invention, accurate short-term traffic flow prediction can be realized, and the prediction precision is superior to that of an existing ODE-based traffic flow prediction model.
本发明公开一种基于深度学习模型的复杂交通网络流量预测方法,包括:根据路网的历史交通数据以及根据谱聚类之后的聚类数据,针对这两种交通历史数据构建动态图;将历史序列通过卷积模型嵌入到潜在维度,对两种不同层次的特征静态融合;将两个层次的特征采用STODE‑Block块来捕捉时间和空间的依赖关系,对时间和空间进行连续的统一建模,得到学习之后的特征表示;对区域和节点的特征表示通过动态融合模块进行第二次动态融合,得到最终的特征输出;采用跳跃连接的方式将两次融合结果进行结合,然后通过ReLU激活函数输出最终的预测结果。本发明能实现精准的短时交通流预测,预测精度优于已有的基于ODE的交通流预测模型。
Complex traffic network flow prediction method based on deep learning model
一种基于深度学习模型的复杂交通网络流量预测方法
2024-10-15
Patent
Electronic Resource
Chinese
IPC: | G08G Anlagen zur Steuerung, Regelung oder Überwachung des Verkehrs , TRAFFIC CONTROL SYSTEMS |
Traffic Flow Prediction Model Based on Deep Learning
British Library Conference Proceedings | 2022
|Traffic flow prediction method based on deep learning nerve network structure
European Patent Office | 2015
|Graph neural network traffic flow prediction method based on deep learning
European Patent Office | 2023
|