CATF framework improves context-aware time-series forecasting
Original title: CATF: A Manager-Worker Framework for Context-Aware Time-Series Forecasting
Manager-Worker architecture with hard routing outperforms standalone models and Time-MoE on five context-rich datasets.
CATF is a Context-Aware Time-Series Forecasting framework built on a Manager-Worker architecture. A Transformer-based Manager routes inputs to Workers using hard routing during training and inference. Workers specialize without manual partitioning. On five datasets, four horizons, and two normalization schemes, CATF improves over its standalone counterpart in 162 of 200 comparisons.
- Hard routing during both training and inference
- Any single-model forecaster can serve as Worker
- Outperforms standalone in 162 of 200 comparisons
- Best CATF-enhanced model outperforms best baseline in 19 of 20 settings
- Outperforms zero-shot Time-MoE in 15 of 20 settings
From the original article
1 min read at arxiv.orgWe study context-aware time-series forecasting, in which a single target variable is predicted from its own history and a set of contextual variables, such as weather and calendar information.
By Yueqiao Chen · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.