arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2604.26668stat.ME
收藏

非线性概率预测协调

Nonlinear Probabilistic Forecast Reconciliation

SUPSI, Institute for Applied Sustainability in the Built Environment(应用建筑环境可持续发展研究所);IDSIA (USI-SUPSI), Dalle Molle Institute for Artifical Intelligence(达勒莫勒人工智能研究所)

Anubhab Biswas, Lorenzo Zambon, Lorenzo Nespoli, Giorgio Corani

AI总结:

本文首次提出处理非线性约束的概率预测协调方法,通过投影和条件化两种方法提升预测准确性,UKF方法在速度和性能上表现最佳。

摘要:

本文首次提出处理非线性约束的概率预测协调方法,通过投影和条件化两种方法提升预测准确性,UKF方法在速度和性能上表现最佳。

英文摘要:

Forecast reconciliation adjusts independently generated forecasts across multiple time series so that they satisfy some known constraints. While probabilistic forecast reconciliation is well established for linear constraints, some practical forecasting problems involve nonlinear relationships among time series. In this paper, we address probabilistic forecast reconciliation with nonlinear constraints for the first time. We extend both reconciliation via projection and conditioning to the case of nonlinear constraints. The former approach reconciles forecast samples by mapping them onto the nonlinear coherent manifold. The latter approach conditions the joint forecast distribution on the constraints; the reconciled distribution is approximated with an algorithm based on the Unscented Kalman Filter (UKF). We evaluate both methods on synthetic and real datasets. Empirically, both reconciliation approaches generally improve forecast accuracy. The UKF-based approach achieves the best overall performance, is consistently retained in the Model Confidence Set -- either alone or alongside other reconciliation methods -- and is substantially faster than the projection-based approach.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑