Characteristic Root Analysis and Regularization for Linear Time Series Forecasting

Researchers present a systematic study of linear models for time series forecasting, focusing on characteristic roots in temporal dynamics and introducing two regularization strategies (Reduced-Rank Regression and Root Purge) to address noise-induced spurious roots. The work achieves state-of-the-art results by combining classical linear systems theory with modern machine learning techniques.

• →Linear models for time series forecasting can be surprisingly competitive with complex models when properly regularized.
• →Characteristic roots govern long-term behavior in time series, but noise tends to produce spurious roots that degrade performance.
• →Two new regularization methods - Reduced-Rank Regression (RRR) and Root Purge - effectively recover low-dimensional latent dynamics.
• →The research demonstrates that mitigating noise influence requires disproportionately large training datasets without proper regularization.
• →Combining classical linear systems theory with modern learning techniques produces robust, interpretable, and data-efficient forecasting models.