Choose timezone
Your profile timezone:
ECAS-ENBIS Course: Adaptive Machine Learning for Time Series Forecasting
→
Europe/Rome
Description
ECAS-ENBIS Course: Adaptive Machine Learning for Time Series Forecasting
Part of the ENBIS-26 Florence conference.
This half-day course is a joint initiative from ENBIS and ECAS (http://ecas.fenstats.eu/) which has provided courses since 1987 to achieve training in special areas of statistics for both researchers and teachers for universities and professionals in industry fields.
Instructor
Yannig Goude (EDF, France)
Overview
This course focuses on adaptive machine learning tools for time series forecasting. Drawing on research developed at EDF R&D, we will cover interpretable machine learning methods, such as Generalized Additive Models (GAMs) and Kalman-filtered GAMs, designed to adapt to non-stationary contexts (e.g., data drift, structural breaks). The second part of the course explores modern approaches, including foundation models for time series and tabular data, as well as online expert aggregation methods.
We will discuss the theoretical foundations of these methods alongside practical examples (R and Python notebooks) using the following packages:
• mgcv (GAMs)
• qgam (Quantile GAMs)
• viking (State-Space Models Inference by Kalman or Viking)
• opera (Online Prediction by Expert Aggregation)
• pytabkit
We will illustrate these methods using real-world datasets (electricity demand, renewable production, electricity prices). These datasets provide excellent examples of time-varying environments, reflecting longer-term changes in consumption habits and the increasing penetration of intermittent power generation.
This 4-hour applicative course will include about 1 hour of hands-on practice (notebook presentation).
Short bio
Yannig Goude is a Senior Researcher at EDF R&D and an Associate Professor in the Mathematics Department at Université Paris-Saclay (France) where he currently teaches courses on machine learning and time series analysis. At EDF R&D, he works at the OSIRIS department (Optimization, Simulation, Risk, and Statistics). His research centers on statistical methods and machine learning for the energy sector, with specific interests in:
• Time series forecasting
• Generalized Additive Models (GAMs)
• Aggregation of Experts
• Application on energy mangement
He has authored numerous papers in leading journals such as IEEE Transactions on Smart Grid and the Journal of the Royal Statistical Society, JASA, NeurIPS, ICML..
Bibliography
• Antoniadis, A., Cugliari, J., Fasiolo, M., Goude, Y., & Poggi, J. M. (2024). Statistical Learning Tools for Electricity Load Forecasting. Springer International Publishing AG.
• Gaillard, P., Goude, Y., & Nedellec, R. (2016). Additive models and robust aggregation for GEFCom2014 probabilistic electric load and electricity price forecasting. International Journal of forecasting, 32(3), 1038-1050.
• Fasiolo, M., Wood, S. N., Zaffran, M., Nedellec, R., & Goude, Y. (2021). Fast calibrated additive quantile regression. Journal of the American Statistical Association, 116(535), 1402-1412.
• De Vilmarest, J., & Goude, Y. (2022). State-space models for online post-covid electricity load forecasting competition. IEEE Open Access Journal of Power and Energy, 9, 192-201.
• Wood, S. N. (2017). Generalized additive models: an introduction with R. chapman and hall/CRC.
• Cesa-Bianchi, N., & Lugosi, G. (2006). Prediction, learning, and games. Cambridge university press.
• Holzmüller, D., Grinsztajn, L., & Steinwart, I. (2024). Better by default: Strong pre-tuned mlps and boosted trees on tabular data. Advances in Neural Information Processing Systems, 37, 26577-26658.