Zero-Shot Time Series Forecasting with Covariates via In-Context Learning
This work addresses the challenge of incorporating covariates effectively in zero-shot forecasting for time series analysis, which is incremental as it builds on existing pretrained models by adding covariate support.
The paper tackles the problem of zero-shot time series forecasting with covariates by introducing COSMIC, a model that uses in-context learning and Informative Covariate Augmentation to train without datasets containing covariates, achieving state-of-the-art performance in zero-shot forecasting with and without covariates.
Pretrained time series models, capable of zero-shot forecasting, have demonstrated significant potential in enhancing both the performance and accessibility of time series forecasting. However, existing pretrained models either do not support covariates or fail to incorporate them effectively. We introduce COSMIC, a zero-shot forecasting model that utilizes covariates via in-context learning. To address the challenge of data scarcity, we propose Informative Covariate Augmentation, which enables the training of COSMIC without requiring any datasets that include covariates. COSMIC achieves state-of-the-art performance in zero-shot forecasting, both with and without covariates. Our quantitative and qualitative analysis demonstrates that COSMIC effectively leverages covariates in zero-shot forecasting.