Root Mean Squared Error (RMSE)
Root Mean Squared Error · Also known as: RMSE, RMS error, quadratic mean error
Root Mean Squared Error is a widely used metric that measures the average magnitude of prediction errors in regression models. Originating from Carl Friedrich Gauss's work on least-squares estimation (1809), RMSE quantifies how far predictions deviate from observed values by averaging the squared differences and taking the square root.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
RMSE is appropriate for continuous regression problems where you want a single-number summary of prediction accuracy. It is especially useful when large errors are particularly costly (e.g., financial forecasting, safety-critical applications). RMSE is sensitive to outliers due to the squaring step, so be aware if your data contains extreme values. Use other metrics like MAE if you want equal penalty for all error magnitudes.
Strengths & limitations
- Interpretable: expressed in the same units as the target variable
- Penalizes larger errors more, aligning with many real-world cost structures
- Smooth and differentiable, making it suitable for optimization algorithms
- Standard across machine learning and statistics, facilitating communication
- Highly sensitive to outliers; a few large errors dramatically inflate RMSE
- Does not directly measure percentage error, making cross-dataset comparison difficult
- Penalizes under- and over-predictions equally, regardless of context
- Can be biased estimator on small samples
Frequently asked
Why take the square root? Why not just use MSE?
The square root returns the metric to the original scale of the target variable, making it interpretable. MSE is in squared units, which is harder to understand. However, MSE is preferred for optimization because it is smooth and differentiable.
How do I compare RMSE values across different datasets?
RMSE is scale-dependent; comparing raw values across datasets with different target ranges is meaningless. Normalize RMSE by dividing by the mean or standard deviation of the target, yielding normalized RMSE (NRMSE) for cross-dataset comparison.
What is a good RMSE value?
There is no universal threshold; it depends on the scale and context of your target variable. RMSE = 5 is excellent for a target with range 1–100, but poor for a target with range 1–1,000. Always benchmark against baseline models and domain expectations.
How does RMSE relate to R²?
Both measure model quality but from different angles. R² quantifies the fraction of variance explained (unitless, 0 to 1), while RMSE measures absolute error magnitude (in original units). A high R² does not guarantee low RMSE if the target has high variance.
Should I use RMSE or MAE?
Use RMSE if large errors are particularly costly; use MAE if you want equal penalty for all error magnitudes. RMSE is more popular but MAE is more robust to outliers. Plot both alongside residuals to gain full insight.
Sources
- Gauss, C. F. (1809). Theoria Motus Corporum Coelestium in Sectionibus Conicis Solem Ambientium. Hamburg: Perthes and Besser. link ↗
- Legendre, A. M. (1805). Nouvelles méthodes pour la détermination des orbites des comètes. Paris: F. Didot. link ↗
- Hastie, T., Tibshirani, R., & Friedman, J. (2009). The Elements of Statistical Learning: Data Mining, Inference, and Prediction (2nd ed.). New York: Springer. DOI: 10.1007/978-0-387-84858-7 ↗
How to cite this page
ScholarGate. (2026, June 3). Root Mean Squared Error. ScholarGate. https://scholargate.app/en/model-evaluation/root-mean-squared-error
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Mean Absolute ErrorModel Evaluation↔ compare
- Mean Absolute Percentage ErrorModel Evaluation↔ compare
- Mean Squared ErrorModel Evaluation↔ compare
- R-squaredModel Evaluation↔ compare