Reliable evaluation of regression models has become a central concern in modern data-driven decision systems, where predictive accuracy must be interpreted alongside robustness, interpretability, and operational feasibility. Despite the widespread adoption of regression algorithms such as Linear Regression, Random Forest Regressor, Support Vector Regression, and Gradient Boosting methods, the selection of appropriate evaluation metrics remains inconsistent across application domains. Metrics including Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Square Error (RMSE), coefficient of determination ($R^{2}$), and Adjusted $R^{2}$ are frequently reported; however, their comparative behavior under varying data characteristics—such as noise, outliers, multicollinearity, and sample size—has not been systematically synthesized in a unified framework. This study presents a structured comparative review of commonly used regression evaluation metrics based on a systematic analysis of peer-reviewed literature and experimental evidence reported across real-world datasets, including housing price prediction, energy consumption forecasting, and healthcare outcome modeling. The review methodology incorporates defined search strategies, inclusion criteria, and cross-domain evidence synthesis to ensure methodological transparency and reproducibility. The analysis highlights that error-based metrics demonstrate strong sensitivity to large deviations and data dispersion, whereas goodness-of-fit measures provide broader insights into model explanatory power but may obscure model complexity effects when multiple predictors are involved. Furthermore, the findings reveal practical trade-offs between interpretability and statistical sensitivity, emphasizing the importance of context-aware metric selection in operational environments. The primary contribution of this work lies in providing a consolidated, application-oriented comparison of regression evaluation metrics that supports researchers and practitioners in selecting appropriate performance measures for reliable and interpretable predictive modeling in real-world scenarios.

Regression Evaluation Metrics, Mean Absolute Error (MAE), Root Mean Square Error (RMSE), R-squared (R²), Adjusted R-squared, Model Performance Evaluation, Predictive Modeling, Machine Learning Evaluation