CAST-CKT: 面向交叉城市流量预测的混沌感知时空-跨城市知识迁移框架
人工智能
2026-02-06 v1 机器学习
摘要
在数据稀缺、跨城市的场景中进行流量预测具有挑战性，因为存在复杂的非线性动力学和 domain shift。现有方法往往无法捕捉流量内在混沌本质，以实现有效的 few-shot learning。我们提出 CAST-CKT，一个 novel Chaos-Aware Spatio-Temporal and Cross-City Knowledge Transfer 框架。它采用高效的混沌分析器来量化流量可预测性制度，驱动 several 关键创新: chaos-aware attention for regime-adaptive temporal modelling; adaptive topology learning for dynamic spatial dependencies; and chaotic consistency-based cross-city alignment for knowledge transfer。该框架还提供具有不确定性量化的 horizon-specific 预测。理论分析显示 improved generalisation bounds。广泛的在四个跨城市 few-shot setting 的实验表明，CAST-CKT 在 MAE 和 RMSE 上均显著优于 state-of-the-art 方法，同时提供可解释的制度分析。代码可在 https://github.com/afofanah/CAST-CKT 获取。
引用
@article{arxiv.2602.05133,
title = {CAST-CKT: Chaos-Aware Spatio-Temporal and Cross-City Knowledge Transfer for Traffic Flow Prediction},
author = {Abdul Joseph Fofanah and Lian Wen and David Chen and Alpha Alimamy Kamara and Zhongyi Zhang},
journal= {arXiv preprint arXiv:2602.05133},
year = {2026}
}