基于动态时空图卷积神经网络的交通预测方法、系统、设备及介质
By decomposing and fusing features of traffic flow data using wavelet transform and deep separable convolutional neural networks, and combining dynamic graph convolution and dilated causal convolutional networks, the problem of low prediction accuracy in existing methods is solved, and higher accuracy traffic flow prediction is achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUNAN UNIV
- Filing Date
- 2023-09-13
- Publication Date
- 2026-07-17
AI Technical Summary
Existing traffic flow prediction methods fail to fully exploit the potential characteristics of traffic flow data, resulting in low prediction accuracy and an inability to accurately describe the dynamic and complex changes in traffic flow in urban road networks.
Traffic flow data is decomposed into approximate and detail components of different frequencies using wavelet transform. Deeply separable convolutional neural networks are used to mine latent features. A dynamic graph is constructed by combining static and adaptive adjacency matrices. Dynamic graph convolution and multilayer dilated causal convolutional networks are used to capture the dynamic spatiotemporal correlation of traffic flow.
It improves the accuracy of traffic flow forecasting, enabling a more accurate description of dynamic and complex traffic flow changes in urban road networks, thus enhancing forecast accuracy.
Smart Images
Figure CN117218837B_ABST