📏 05 — Regression Metrics¶

Regression metrics measure prediction error.

MAE¶

Mean Absolute Error:

from sklearn.metrics import mean_absolute_error

mae = mean_absolute_error(y_test, y_pred)

Easy to interpret: average absolute mistake.

from sklearn.metrics import mean_squared_error

mse = mean_squared_error(y_test, y_pred)
rmse = mean_squared_error(y_test, y_pred, squared=False)

RMSE penalizes large errors more than MAE.

from sklearn.metrics import r2_score

r2 = r2_score(y_test, y_pred)

R² measures how much variance is explained by the model. Higher is better, but it can be misleading alone.

residuals = y_test - y_pred

Plot residuals to inspect patterns:

import matplotlib.pyplot as plt

plt.scatter(y_pred, residuals)
plt.axhline(0, color="red")
plt.show()

Random-looking residuals are better than patterned residuals.

Metric	Use When
MAE	you want easy interpretation
RMSE	large errors are especially bad
R²	you want variance explained