DigiNews

Tech Watch by Johan Denoyer

← Back to articles

The Unreasonable Difficulty of Time Series Forecasting

Quality: 8/10 Relevance: 9/10

Summary

The article argues that time series forecasting remains difficult due to the data generating process, low signal-to-noise, data scarcity, and distribution shift. It compares simple baselines and foundation models, highlighting that in many cases straightforward methods outperform complex models on seasonally strong data, and discusses the role of exogenous features for real-world forecasting.

🚀 Service construit par Johan Denoyer