Timersi improves time-series forecasting pipelines with self-correction
TIMERSI: Towards Self-Improved Agentic Deployment for Time-Series Foundation Models
Machine Learning
Summary
Time-series models predict future data points but often need specific adjustments to work well in different real-world cases. The authors present TIMERSI, a system that automatically improves these models by making small changes based on past errors and testing those changes. It picks the best adjustment for current conditions, leading to more accurate forecasts without needing experts to redesign models for each task. Their tests on many forecasting problems show TIMERSI beats other methods in accuracy.
What this means in practice
- •For data science teams: Improve forecasting accuracy by automatically refining model input-output pipelines for diverse time-series tasks without manual redesign.
- •For energy demand planners: Enhance energy usage forecasts by deploying adaptive pipelines that adjust to varying conditions to reduce recurring forecast errors.
Authors
Haoxin Liu, Deqing Fu, Lucheng Fu, Yiqiao Jin, Rushi Qiang, Harshavardhan Kamarthi, B. Aditya Prakash
Abstract
Time-series foundation models provide strong zero-shot forecast, but deployment for real- world applications still requires task-specific designs for model inputs and outputs. Direct inference can leave domain knowledge and re- curring forecast errors unaddressed, while tai- lored pipelines require substantial human ex- pertise and effort. We introduce TIMERSI, a self-improving agentic deployment system around frozen forecasting models. A cod- ing agent translates task descriptions into ex- ecutable pipeline changes. It tests each re- vision on historical forecasting windows and uses observed errors to guide subsequent re- visions. During deployment, an adaptive se- lector chooses among the resulting pipelines using completed forecasts under similar con- ditions. Across 198 tasks on TIME and FEV- bench, TIMERSI outperforms all submissions on six point and probabilistic forecasting met- rics. Controlled comparisons support the value of feedback-guided revision over non-iterative pipeline search. Further experiments verify the generalization with different coding agents and forecasting models.