一种基于动态随机自注意力网络的交通流量预测方法
By using a dynamic stochastic self-attention network, spatiotemporal information is embedded in the traffic flow tensor and the road network adjacency matrix to learn spatial and temporal attention matrices, thus solving the problems of high computational complexity and insufficient accuracy of existing methods and achieving efficient traffic flow prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHEJIANG UNIV OF TECH
- Filing Date
- 2026-05-15
- Publication Date
- 2026-07-17
AI Technical Summary
Existing traffic flow prediction methods based on attention mechanisms have high computational complexity and fail to fully utilize the temporal periodicity of traffic data, resulting in insufficient accuracy in spatial modeling.
A dynamic stochastic self-attention network is adopted. By constructing a traffic flow tensor and a road network adjacency matrix, spatiotemporal information is embedded, spatial and temporal attention matrices are learned, and multi-level iterations are performed to fuse spatiotemporal feature representations and finally map future traffic flow predictions.
While reducing computational complexity, it improves the spatial modeling accuracy of traffic flow prediction, fully explores spatiotemporal dependencies and periodic patterns, and enhances prediction accuracy.
Smart Images
Figure CN122201005B_ABST