Signature
PrecisionRecallDisplay.from_estimator(estimator, X, y)What it does
PrecisionRecallDisplay shows the tradeoff between finding positives and avoiding false alarms. The baseline is the positive rate, which makes it more informative than ROC under heavy class imbalance.
Parameters
estimatorestimator- A fitted classifier with predict_proba or decision_function.
Xarray-like- Feature matrix to score.
yarray-like- True binary labels.
Returns
PrecisionRecallDisplay
Example
pythonPrecisionRecallDisplay example
import matplotlib.pyplot as plt
from sklearn.metrics import PrecisionRecallDisplay
PrecisionRecallDisplay.from_estimator(model, X_test, y_test)
plt.show()Watch out for
- The no-skill baseline equals the positive class rate, not 0.5.
- Precision at one threshold does not describe the whole curve.