Forecast combination method controls risks and improves accuracy
Downside-Controlled Online Forecast Combination under Delayed and Revised Outcomes
Machine Learning
Summary
Forecasting models often need to be corrected after they make predictions, especially when outcomes arrive later or are updated. The authors propose a way to combine a fixed forecaster, a static correction, and an online correction so that the combined forecast doesn’t perform much worse than the original. Their method keeps risks low while improving accuracy, tested on energy load data and other benchmarks. They also identify when their approach works best based on factors like how quickly experts update predictions and how outcomes align.
What this means in practice
- •For energy forecasters: Improve day-ahead electricity load predictions across multiple zones with controlled accuracy loss and potential gains.
- •For financial risk teams: Combine legacy and adaptive forecasts while limiting downside losses when outcomes are delayed or revised.
Authors
Minkyoung Kim, Hyunjung Byun, Yohan Lee, Beakcheol Jang
Abstract
Post-hoc correction adjusts a forecaster that cannot be retrained, such as a foundation model, but a correction fitted where errors are stable can hurt where they shift. We aim for downside control: not much worse than the starting forecast. We combine the frozen forecaster, a static corrector and an online corrector on the simplex, using only losses that mature after the horizon. Across seven benchmarks and four base models, two of them foundation models, the worst deterioration over 28 pairs at the main horizon is 0.15% and gains reach 11.5%. On day-ahead load for seven European bidding zones it lowers mean MSE in all seven zones, while single correctors raise mean MSE by up to 102% where the published forecast is most accurate. Three empirical conditions on expert speed, stream length and outcome alignment, each fixed by a documented failure, delimit its scope. Learning from the provisional outcome improves four zones on the settled one; learning on the settled outcome restores all seven.