This library contains the Python implementation of Venn-ABERS calibration for binary and multiclass classification problems.
pip install venn-abers
The method can be applied on top of an underlying scikit-learn algorithm.
from sklearn.datasets import make_classification
from sklearn.model_selection import train_test_split
from sklearn.naive_bayes import GaussianNB
from venn_abers import VennAbersCalibrator
X, y = make_classification(n_samples=1000, n_classes=3, n_informative=10)
X_train, X_test, y_train, y_test = train_test_split(X, y)
clf = GaussianNB()
# Define Venn-ABERS calibrator
va = VennAbersCalibrator(estimator=clf, inductive=True, cal_size=0.2, random_state=101)
# Fit on the training set
va.fit(X_train, y_train)
# Generate probabilities and class predictions on the test set
p_prime = va.predict_proba(X_test)
y_pred = va.predict(X_test)
Further examples can be found in the github repository https://github.com/ip200/venn-abers in the examples
folder:
If you find this library useful please consider citing: