pcalibration 0.1.2

No project description provided

probability-calibration

Probability calibration is a technique to calibrate the probabilities of a classifier. This package provides a simple interface to calibrate the probabilities of a sklearn-stype classifier. The package also provides a set of calibration methods.

Methods

• Platt Scaling
• Isotonic Regression
• Inductive Venn-Abers Predictor (IVAP)
• Cross Venn-Abers Predictor (CVAP)

Installation

$ pip install pcalibration

Usage

from pcalibration import CalibratorWrapper, CrossVennAbers

# Load the data
X_train, X_test, y_train, y_test = ...

# Create a CalibratorWrapper object
calibrator = CalibratorWrapper(model, CrossVennAbers(k=5))

# Fit the calibrator
calibrator.fit(X_train, y_train)

# Predict the probabilities
y_pred = calibrator.predict_proba(X_test)