SciKit Learn can visualize how well tuned a classifier is via [the `calibration` package](https://scikit-learn.org/stable/auto_examples/calibration/plot_calibration_curve.html#sphx-glr-auto-examples-calibration-plot-calibration-curve-py). ![[Calibration Plots.png]] # `predict_proba` and `decision_function` Calibration displays can be created for classifiers with their own `predict_proba` method. While classifiers that support the `decision_function` method can be [extended with a hand-rolled](https://scikit-learn.org/stable/auto_examples/calibration/plot_calibration_curve.html#linear-support-vector-classifier) `predict_proba` method. Finally, wrapping another classifier via a regressor, either a [sigmoid](https://scikit-learn.org/stable/modules/calibration.html#sigmoid) (default) or [isotonic](https://scikit-learn.org/stable/modules/calibration.html#isotonic), can provide the needed prediction probability and/or enhance the classifier output to have more meaningful prediction probabilities. This enhancement is achieved by the [`CalibratedClassifierCV`](https://scikit-learn.org/stable/modules/generated/sklearn.calibration.CalibratedClassifierCV.html#sklearn.calibration.CalibratedClassifierCV) class.