The research assesses how four regression models namely Lasso, Ridge, Elastic Net and Best Subset Model perform at economy variable predictions with emphasis on prediction precision and overall stability along with generalization ability. The main evaluation metrics comprise of Mean Squared Error (MSE) and R-Squared together with Adjusted R-Squared and Leave-One-Out Cross-Validation (LOOCV) MSE. Results demonstrate that the Best Subset Model stands out with its exceptional accuracy through MSE measurements of 0.001 and 0.0016 in LOOCV. Alternatively, Ridge Regression shows strength in dealing with multicollinearity problems. Elastic Net maintains a balanced combination of feature selection capabilities with multi-collinearity control while Lasso demonstrates strong feature selection yet it leads to overly-shrunken results. The Best Subset Model stands out as the most dependable predictor for macroeconomic forecasting because it offers the best combination of model complexity and interpretability.
Keywords: Shrinkage regression Models, Macroeconomic Variables, Forecasting, Multicollinearity, Evaluation Metrics.