A Survey of Deep Learning and Foundation Models for Time Series Forecasting

• DOI: 10.48550/arxiv.2401.13912
• Publication Date: 2024-01-01
• Authors: John A. Miller; Mohammed S. Al‐Do...; Farah Saeed; Nasid Habib Barna; Subas Rana; I. Budak Arpinar; Ninghao Liu
• Keywords: Interpretability

Deep Learning has been successfully applied to many application domains, yet its advantages have slow emerge for time series forecasting. For example, in the well-known Makridakis (M) Competitions, hybrids of traditional statistical or machine learning techniques only recently become top performers. With recent architectural advances deep being forecasting (e.g., encoder-decoders with attention, transformers, and graph neural networks), begun show significant advantages. Still, area pandemic prediction, there remain challenges models: is not long enough effective training, unawareness accumulated scientific knowledge, interpretability model. To this end, development foundation models (large extensive pre-training) allows understand patterns acquire knowledge that can be new related problems before training data becomes available. Furthermore, a vast amount available tap into, including Knowledge Graphs Large Language Models fine-tuned domain knowledge. There ongoing research examining how utilize inject such into models. In survey, several state-of-the-art modeling are reviewed, suggestions further work provided.