Professor Rommert Dekker from Erasmus School of Economics and Christos Tzagkarakis (FORTH-ICS institute in Heraklion, Crete) have been awarded the Best Paper Award during the 14th EETN Conference on Artificial Intelligence (SETN) from 9 to 11 September. They received this award in Chania, Crete, for their paper on conformal calibration on probabilistic electricity price forecasts.
Forecasting electricity prices during market shocks
Their winning article examines how different types of AI forecasting models handle sudden economic shocks when paired with CQR (Conformalised Quantile Regression), a distribution-free wrapper designed to guarantee reliable prediction intervals through split conformal calibration. They looked at electricity prices in the Netherlands on the Dutch EPEX market across six different time periods between 2018 and 2024, spanning calm times and the 2021–2022 energy crisis.
The results explore how, while CQR seems to work well in stable markets, it causes AI models to fail during market shocks. The researchers reveal that simpler linear forecasting models are significantly more stable during structural changes, which offers crucial insight for risk management, specifically in the energy market.
The EETN Conference on Artificial Intelligence
Established in 1996 by the Hellenic Artificial Intelligence Society (EETN), the EETN Conference on Artificial Intelligence is one of the most important forums in Greece for Greek and international AI scientists to present original, high-quality research in AI.
About TwinODIS
This research was developed within the framework of the TwinODIS project, funded by the European Union with the goal of combining great minds within the data-intensive field. This project collaboration stands out as a key achievement of the project programme, representing a genuine collaboration between two TwinODIS partners and offering the opportunity for Dekker and Tzagkarakis to combine their expertise.
- Professor
- Related content
