Modèles de Fondation et Ajustement : Vers une Nouvelle Génération de Modèles pour la Prévision des Séries Temporelles
This work addresses time series forecasting challenges for researchers and practitioners, but it is incremental as it reviews existing methods and studies fine-tuning.
The paper tackles the problem of zero-shot time series forecasting by reviewing foundation models and studying fine-tuning effects, finding that fine-tuning generally improves performance, particularly for long-term horizons.
Inspired by recent advances in large language models, foundation models have been developed for zero-shot time series forecasting, enabling prediction on datasets unseen during pretraining. These large-scale models, trained on vast collections of time series, learn generalizable representations for both point and probabilistic forecasting, reducing the need for task-specific architectures and manual tuning. In this work, we review the main architectures, pretraining strategies, and optimization methods used in such models, and study the effect of fine-tuning after pretraining to enhance their performance on specific datasets. Our empirical results show that fine-tuning generally improves zero-shot forecasting capabilities, especially for long-term horizons.