PILOT:基于预测误差的在线时间序列预测按需重训练框架
Pseudo-Label-Triggered Retraining from Forecast Errors for Online Time Series Forecasting
AI 导读
PILOT(Pseudo-label-Informed Learned Online Trigger)是一个在线重训练框架,利用已实现的预测误差作为直接部署反馈,学习何时触发重训练,以伪标签训练轻量评分器。
正文 · 原文
Abstract:Real-world time series forecasting systems operate under non-stationary data streams, where forecasting performance may degrade over time. Although retraining can recover the performance, it incurs non-trivial computational and operational costs. Under limited deployment resources, the key challenge is therefore not only how to retrain but also when to retrain. While existing retraining policies often rely on indirect indicators such as drift alarms or model staleness, we instead use realized forecast errors as direct deployment feedback. In this paper, we propose PILOT (Pseudo-label-Informed Learned Online Trigger), an online retraining framework that learns when to retrain from forecast-error dynamics. Since ground-truth retraining labels are unavailable, PILOT constructs a pseudo-label from future increases in forecast error and trains a lightweight scorer to predict it from observed error states. At deployment, PILOT uses only completed forecast errors and serves as a plug-in module for arbitrary forecasting backbones without architectural modification. We evaluate PILOT under standard multivariate forecasting settings across eight benchmarks with three representative backbones---DLinear, iTransformer, and TimesNet. Across all three backbones, PILOT achieves state-of-the-art average-rank performance among retraining policies while maintaining a favorable performance--efficiency trade-off.
| Comments: | accepted at WSDM 2027 (Main) |
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2609.39789 [cs.LG] |
| (or arXiv:2609.39789v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2609.39789 arXiv-issued DOI via DataCite |
|
| Related DOI: | https://doi.org/10.1145/3862322.3863553
DOI(s) linking to related resources |
Submission history
From: Yoo-Min Jung [view email]
[v1]
Wed, 30 Sep 2026 14:10:20 UTC (381 KB)
[v2]
Tue, 17 Nov 2026 13:41:37 UTC (557 KB)
来源:arXiv cs.AI · arxiv.org