Test-time method improves multivariate time series forecasting accuracy
Correction-space Cross-variate Interaction for Test-time Adaptation in Time Series Forecasting
Machine Learning
Summary
Time series forecasting involves predicting future data points based on past observations. Sometimes the data patterns change over time, making predictions less accurate. The authors found that adjusting the corrections to predictions by considering interactions between different data variables, rather than changing each separately, helps improve accuracy. Their method, called CoRe, smartly refines corrections using shared reference points and adapts them based on new data to reduce errors, especially for longer-term forecasts.
What this means in practice
- •For financial analysts: Enhance stock and economic forecasts by adapting models at prediction time to better handle changing market conditions using cross-variable corrections.
- •For energy system operators: Improve electricity demand forecasting by refining multi-sensor data predictions during operations to maintain accuracy amid distribution shifts.
Authors
Yuanyuan Deng, Mykola Pechenizkiy, Songgaojun Deng
Abstract
Test-time adaptation (TTA) is a promising paradigm for handling distribution shift in time-series forecasting (TSF), where models adapt at inference time, often leveraging delayed observed data to refine predictions. In the multivariate setting, distribution shifts often exhibit cross-variate dependencies, yet existing TSF-TTA methods adapt each variate independently and ignore this cross-variate structure. Exploiting such structure motivates cross-variate interaction, but coupling variates through backbone predictions introduces direct pathways for mixing uncorrected errors across variates, a concern under the delayed supervision of TSF-TTA. We identify the \emph{interaction space} as a key design choice, and show that acting on adapter corrections that refine backbone outputs, the \emph{correction space}, rather than on the predictions themselves, avoids directly propagating backbone errors across variates. We build on this to propose \textsc{CoRe} (\textsc{Co}rrection-space Interaction \textsc{Re}finement), realizing correction-space interaction through (i) Shared-anchor Correction Refinement (SCR), which combines each variate's correction with a shared anchor through a parameter-efficient bottleneck, and (ii) input-conditioned spectral gating, which adaptively modulates the refinement from the current input window. Across seven backbones, six datasets, and four prediction horizons, \textsc{CoRe} reduces MSE by 25.82\% on average over backbones and 10.57\% over the state-of-the-art TSF-TTA method, with stronger gains at medium-to-long horizons and modest computational overhead. Data and code are available at: https://github.com/yyddou/CoReTTA