Forget Before You Connect: Pre-Graph Subspace Projection for Channel-Level Unlearning in Multivariate Time Series

Machine unlearning has largely focused on removing training samples or classes, leaving one question underexplored: *How can a deployed model reduce its dependence on an input channel?* In multivariate time-series (MTS) systems, corrupted or privacy-sensitive sensors may need to be removed after deployment. Existing approaches require costly retraining or iterative optimization, while permanent input masking can degrade predictive utility without addressing learned cross-channel dependencies. We introduce *Channel-Centric Subspace Projection* (CCSP), a training-free method that identifies target-associated latent directions via singular value decomposition and attenuates them before cross-channel graph propagation. An optional trainable utility-repair extension is evaluated separately. Across three MTS classification benchmarks, CCSP preserves near-original utility with low residual sensitivity, achieving $2.56\sim 48.80\times$ speedup over iterative baselines. On UCI-HAR, it achieves a $40.98\%$ mean sensor-wise reduction in linear intervention leakage and reduces mean prediction disagreement from $9.54\%$ to $8.12\%$. On C-MAPSS remaining-useful-life regression, CCSP further reduces target-induced prediction variation by $31.41\%$ on average across $14$ sensors while maintaining comparable predictive utility. These results establish latent subspace intervention as an efficient approach to channel-level forgetting, while distinguishing approximate influence removal from certified information erasure.