JudgeCast improves time series forecasting using experience and covariate judgments
JudgeCast: Time Series Forecasting with Experience-Informed Covariate Judgements
Machine LearningArtificial Intelligence
Summary
Forecasting future events often depends on various factors, called covariates, which can change over time and context. The authors created JudgeCast, a method that first makes a basic forecast and then adjusts it by carefully judging each factor's effect, using past forecast errors to improve future predictions. This approach helps the system learn which judgments worked best in the past, leading to more accurate forecasts. Tests on real-world data showed that JudgeCast performs better than other strong forecasting methods.
What this means in practice
- •For demand forecasting teams: Improve sales predictions by adjusting forecasts using explicit judgments of influencing factors and learning from past forecast errors.
- •For energy grid operators: Enhance electricity demand forecasting by incorporating learned judgments on weather and usage covariates to better adapt to changing conditions.
Authors
Donguk Kwon, Wooseok Jeong, Dongha Lee
Abstract
Covariate effects vary across contexts and shift over time, requiring forecasters to assess how to use them for each forecasting context. As forecasting proceeds, observations for earlier forecasts become available, providing feedback on past covariate use for subsequent forecasts. However, when multiple covariates act together, the forecast error reveals the numerical discrepancy from the observation but not how the covariates should have been used. We introduce JudgeCast, an experience-based framework for time series forecasting with covariates. Following the judgmental adjustment practice, a frozen TSFM provides the base forecast, while a frozen LLM uses the current context and relevant experience to adjust it. Within the adjustment, assessing covariate effects and determining the numerical adjustment serve distinct roles, so JudgeCast first forms explicit covariate-wise judgments and then determines the adjustment. After observation, JudgeCast uses the observed residual of the base forecast to reconstruct alternative judgments and evaluates the original and alternatives through their resulting adjustments. The best-performing decision is selected and retained as validated experience for subsequent forecasts. Across diverse real-world datasets, JudgeCast outperforms strong baselines. Ablations show that explicit covariate-wise judgment can improve forecast-time adjustment, while residual-guided experience construction yields more reliable forecasting gains than retaining raw decisions as experience.