Traffic situation prediction method, system and device based on spatial-temporal feature enhancement and medium
By constructing a traffic situation prediction method with enhanced spatiotemporal features, utilizing multi-source data and shock wave propagation mechanism, and combining gradient boosting tree model and cloud-edge collaboration mechanism, the problem of insufficient model adaptability in existing technologies is solved, and efficient, real-time early warning and error correction capabilities for traffic situations are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- ZHEJIANG UNIV OF TECH
- Filing Date
- 2026-04-09
- Publication Date
- 2026-05-15
AI Technical Summary
Existing traffic situation prediction methods lack online adaptive capabilities and cannot self-correct using real traffic evolution. This makes the models prone to underreporting when faced with long-tailed distributed samples, and they cannot actively capture the propagation process of congestion shock waves.
A traffic situation prediction method based on spatiotemporal feature enhancement is constructed. An initial spatiotemporal feature tensor is built by acquiring multi-source traffic flow data, the shock wave spatial evolution term is fused, a gradient boosting tree model is used for prediction, and the model parameters and structure are dynamically updated through a cloud-edge collaborative online adaptive evolution mechanism. The method is combined with a simulation sample expansion mechanism to correct missed reports.
It significantly improves the model's ability to perceive complex traffic evolution processes and the interpretability of early warning results, enhances its ability to respond to emergencies, and achieves a balance between real-time performance, temporal integrity, and adaptability.
Smart Images
Figure CN122050153A_ABST