Classification quickstart

Run calibrated explanations for binary or multiclass classification.

Prerequisites

pip install calibrated-explanations scikit-learn

Classification semantics note

• Calibration prerequisites: fit on x_proper, y_proper and calibrate on held-out x_cal, y_cal.

• Mode-specific guarantees: Venn-Abers provides calibrated probability intervals for class predictions.

• Assumptions: calibration and deployment data are exchangeable or distribution-matched.

• Explicit non-guarantees: no guarantee under drift or regime shift.

• Explanation-envelope limits: feature-level intervals are model-behavior summaries, not causal guarantees.

• Formal semantics: Calibrated interval semantics.

1. Load data and split sets

from sklearn.datasets import load_breast_cancer
from sklearn.model_selection import train_test_split

dataset = load_breast_cancer()
x = dataset.data
y = dataset.target

x_train, x_test, y_train, y_test = train_test_split(
    x, y, test_size=0.2, stratify=y, random_state=0
)
x_proper, x_cal, y_proper, y_cal = train_test_split(
    x_train, y_train, test_size=0.25, stratify=y_train, random_state=0
)

2. Fit and calibrate the explainer

from sklearn.ensemble import RandomForestClassifier
from calibrated_explanations import WrapCalibratedExplainer

explainer = WrapCalibratedExplainer(RandomForestClassifier(random_state=0))
explainer.fit(x_proper, y_proper)
explainer.calibrate(x_cal, y_cal, feature_names=dataset.feature_names)

3. Generate calibrated factual explanations

factual = explainer.explain_factual(x_test[:5])

For multiclass all-class explanations:

multi_factual = explainer.explain_factual(x_test[:5], multi_labels_enabled=True)

4. Explore calibrated alternatives

alternatives = explainer.explore_alternatives(x_test[:2])

Next steps:

• Interpret calibrated explanations

• Alternative explanations and ensured filters

• Citing calibrated-explanations

Entry-point tier: Tier 2.