Ctrl+K

R2 Score

R2 Score#

Overview#

The R² score, also known as the coefficient of determination, is a statistical measure that represents the proportion of the variance for the dependent variable that’s explained by the independent variables in a regression model. It is a key indicator of the model’s goodness of fit.

The R² score is defined as:

\[ R^2 = 1 - \frac{\sum_{i=1}^{n} (y_i - \hat{y}_i)^2}{\sum_{i=1}^{n} (y_i - \bar{y})^2} \]

where:

  • \( y_i \) are the actual values,

  • \( \hat{y}_i \) are the predicted values,

  • \( \bar{y} \) is the mean of the actual values,

  • \( n \) is the number of data points.

Interpretation#

  • R² = 1: The model perfectly explains the variance in the data.

  • R² = 0: The model does not explain any of the variance in the data.

  • R² < 0: The model is worse than a horizontal line (mean of the target values).

Real-World Example#

Consider you’re developing a model to predict house prices based on features like size, number of bedrooms, age, and location. After training your model, you calculate the R² score to understand how well your model captures the variability in house prices.

Example#

Here’s how you can calculate the R² score using Python and scikit-learn, assuming you already have a trained model and test data.

import numpy as np
from sklearn.metrics import r2_score

# Assume y_test and y_pred are the actual and predicted values from the test set
y_test = np.array([300000, 450000, 200000, 500000, 700000])
y_pred = np.array([310000, 430000, 210000, 480000, 690000])

# Calculate R² score
r2 = r2_score(y_test, y_pred)
print(f"R² Score: {r2}")

Practical Use#

The R² score helps in comparing different models. For instance, you might want to compare a Ridge Regression model and a Lasso Regression model to determine which one better explains the variance in house prices. While a higher R² score generally indicates a better fit, it’s essential to be cautious of overfitting, where the model performs exceptionally well on training data but poorly on unseen data.

Limitations#

  1. Does not indicate model accuracy: A high R² score doesn’t mean the model is accurate; it just means the model explains a large portion of the variance.

  2. Sensitive to outliers: Outliers can significantly affect the R² score, potentially leading to misleading conclusions.

  3. Not suitable for non-linear models: R² score assumes a linear relationship between the dependent and independent variables. For non-linear models, other metrics might be more appropriate.

Conclusion#

The R² score is a valuable metric for evaluating the performance of a regression model, indicating how well the model explains the variance in the target variable. However, it should be used in conjunction with other evaluation metrics to ensure a comprehensive assessment of model performance.

Notes#

for normalized MSE = 1 => the prediction is as good as the mean of the target.

we want MSE <1 ;

  • R2 = 0 good as the mean of the target.

  • R2 = 1 perfect prediction.

  • R2 < 0 worse than the mean of the target.

in time series - forcating - use the mean may be a very good prediction.

Contents