DataTerminology.comNEW
The scikit-learn regression score measuring variance explained by a model relative to a constant baseline.
DESCRIPTION
R2 score in ML toolkits operationalizes the coefficient of determination for evaluating regression models. A score of 1.0 indicates perfect predictions; 0.0 equals the mean baseline; negative values indicate underperformance. It is sensitive to outliers and can be inflated by overfitting. Practitioners compare models using cross-validated R2 and complement it with MAE/RMSE.
ALTERNATIVE DEFINITIONS
scikit-learn
R^2 (coefficient of determination) regression score function.
Defines R2 as a concrete scoring function in the library.
Wikipedia
In statistics, the coefficient of determination, denoted R2, is a measure of the proportion of variance in the dependent variable that is predictable from the independent variables.
Provides the statistical foundation for the score.
RELATED TERMS
R-squared Mean Absolute Error RMSE Baseline Model Cross-Validation Regression Metrics scikit-learn Residuals
COMPETING TERMS
Adjusted R-squared Explained Variance Score
RELATIONS
- Broader: Score
KNOWLEDGE AREAS
Data Science AI Mathematics Testing & Validation
TARGET AUDIENCE
Data Scientist MLOps Engineer AI Engineer BI Specialist Data Analyst
EXTERNAL SOURCES
- Wikipedia - Coefficient of determination
- scikit-learn - Official R2 scoring documentation
- statsmodels - Regression diagnostics context
TOP QUESTIONS
Why can the R2 Score be negative?
It becomes negative when the model predicts worse than a constant mean baseline for the target variable.
Category: Misunderstanding
How to use R2 Score during model selection?
Compare cross-validated R2 across candidates and inspect residuals; avoid picking models solely on in-sample R2.
Category: Best practices
What is the relationship between R2 Score and Explained Variance?
They are related but not identical; explained variance measures dispersion of errors, while R2 compares to a baseline.
Category: Misunderstanding
Should I optimize models directly for R2 Score?
Often you optimize for loss (e.g., MSE) and report R2. Direct optimization for R2 is uncommon.
Category: Opinion
How does scaling features impact R2 Score?
Scaling does not change R2 in linear regression, but it can affect optimization convergence and regularization behavior.
Category: Requirements
How to interpret R2 with heteroscedastic data?
Use diagnostics and robust metrics; R2 alone may mask unequal error variance.
Category: Risks
Is R2 suitable for non-linear models?
Yes as a general fit measure, but complement it with residual plots and domain-specific metrics.
Category: Opinion
What baseline does R2 Score use?
A constant predictor equal to the mean of the observed targets for the evaluation set.
Category: Example
How do outliers affect R2 Score?
Outliers can skew fit and inflate or deflate R2. Consider robust modeling or outlier treatment.
Category: Risks