Time series model improves forecasting with multivariate context
$t_0$: A Time-Series Foundation Model for Forecasting with Context
Machine LearningArtificial Intelligence
Summary
Forecasting future events based on past data and related factors is tricky, especially when many variables interact. The authors introduce t0, a type of AI model that looks at multiple time series and context variables together to better predict what comes next. These models don’t need extra training for each specific task and can produce probability-based forecasts, meaning they provide different possible outcomes with likelihoods. Tests show t0 performs well on many benchmark datasets, including electricity demand and price forecasting. The models also handle missing data and long-term predictions effectively.
What this means in practice
- •For energy demand planners: Improve short- and long-term electricity demand forecasts using multivariate context for better grid management.
- •For financial trading analysts: Enhance price prediction accuracy for hourly electricity market prices by reducing error compared to simple lag-based models.
Authors
Lucas Meyer, Claudio Sole, Huikan Xiang, Nicolas Li, Lucas Franceschino, Arnau Quera-Bofarull, Maarten P. Scholl, Joachim Fainberg, Geoffrey Négiar
Abstract
We present $t_0$, a family of open-weights foundation models for forecasting with multivariate context. We release its first two members: $\texttt{t0-alpha}$ and $\texttt{t0-beta}$, respectively 102M and 256M parameters. Both condition their forecasts on target history, past covariates, and known-future covariates, without task-specific retraining. Their transformer layers alternate attention along time and across variates. They produce probabilistic forecasts through quantile predictions. Pretraining combines curated public data with synthetic generator families constructed to contain covariate-to-target dependencies. On GIFT-Eval, $\texttt{t0-alpha}$ reaches an aggregate CRPS of 0.4941, and $\texttt{t0-beta}$ a CRPS of 0.4738 and a MASE of 0.6865, third on both and within 4.0% of the best zero-shot TSFM. On fev-bench they score 42.2 and 46.7 in skill, the latter third again and 2.0 points behind the leader. We analyze $\texttt{t0-alpha}$ in depth. Known-future covariates raise its skill by 6.3 percentage points across 30 tasks. The report also examines its calibration, its rollout strategy on long horizons, and its robustness to missing data. On the Victoria electricity-demand benchmark, $\texttt{t0-beta}$ is among the most accurate models with a context of nearly a year. In an independent Macrocosm evaluation of hourly ERCOT prices over 29 months, both cut the MAE of the lagged-price baseline by 38%.