netcal.binning

Binning Methods for Confidence Calibration

This package consists of several methods for confidence calibration which use binning to approximate confidence estimates to the measured accuracy. The most common calibration methods within this package are netcal.binning.HistogramBinning and netcal.binning.IsotonicRegression.

Available classes

HistogramBinning([bins, equal_intervals, ...])	Simple Histogram Binning calibration method.
IsotonicRegression([detection, ...])	Isotonic Regression method.
BBQ([score_function, equal_intervals, ...])	Bayesian Binning into Quantiles (BBQ).
ENIR([score_function, quick_init, ...])	Ensemble of Near Isotonic Regression (ENIR) models.