Innovation
Methods, ideas, or system contributions that make the work stand out.
Cross-channel importance
Horizon-adaptive source selection
Multivariate time-series forecasting
Functional interventions
Channel-independent forecaster
Rethinking Cross-Channel Importance in Time-Series Forecasting
🤖 AI Summary
This study addresses the conceptual conflation of “statistical correlation,” “predictive utility,” and “model utilization” among channels in multivariate time series forecasting. It formally delineates these three notions, proves their non-equivalence, and reveals the dynamic evolution of cross-channel dependencies across prediction horizons. Methodologically, this work introduces a mechanism- and problem-aware perspective for evaluating relative channel importance and designs a bounded post-hoc support strategy to augment frozen channel-independent predictors. Experiments employ controlled Ridge regression, iTransformer, and TimesNet models to conduct functional interventions and comparative validations. Results demonstrate that adaptive source selection yields an average improvement of 5.16% in linear models, while the post-hoc support strategy proves effective in 12 out of 16 configurations; however, performance gains remain limited for nonlinear architectures.
📝 Abstract
Cross-channel modeling is central to multivariate time-series forecasting, yet channels that are statistically related, predictively useful, and actually used by a trained forecaster are often treated as if they defined the same notion of importance. We show that they need not coincide. Cross-channel dependency structures change substantially across future offsets, and horizon-adaptive source selection improves a controlled Ridge predictor in 21 of 32 dataset--prediction-length conditions, with a mean gain of $5.16\%$. This selected-set signal also transfers to a matched nonlinear predictor. Yet imposing the same horizon-specific source logic on iTransformer yields only 11 of 20 wins and a mean gain of $0.208\%$, with little alignment between controlled and neural gains. Functional interventions further show that strong forecasters use cross-channel information, while their source-reliance rankings agree little with controlled utility or with one another across iTransformer, TimesNet, and a cross-channel TimeMixer. As a constructive consequence, bounded post-hoc support improves a frozen channel-independent forecaster in 12 of 16 dataset--horizon conditions, with a positive aggregate bootstrap interval. Cross-channel importance should therefore be interpreted relative to the forecasting mechanism and question that define it: related $\neq$ useful $\neq$ used.
🔎 Similar Papers
No similar papers found.
