Classification

Use KFCClassifier for categorical labels.

from kfc_procedure import KFCClassifier

clf = KFCClassifier(
    divergences=["euclidean"],
    local_model="decision_tree_classifier",
    combiner="majority_vote",
    n_clusters=2,
    random_state=42,
)
clf.fit(X_train, y_train)
y_pred = clf.predict(X_test)

Classification combiners

Combiner	Behavior
majority_vote	most frequent local prediction
stacking_classifier	logistic regression meta-classifier
combined_classifier	COBRA-style weighted vote

One-class clusters

If a cluster contains only one class, some local classifiers may fail. Reduce n_clusters, use more data, or choose a classifier that can handle small local samples.

Current predict_proba limitation

The inspected code defines KFCProcedure.predict_proba(), but FStep does not currently implement predict_proba(). Probability prediction may require a patch.