calibration_curve#
- EstimatorReport.inspection.calibration_curve(*, data_source='test', n_bins='auto', strategy='quantile')[source]#
Display the calibration curve.
This method is available for binary and multiclass classification reports. It computes one-vs-rest calibration curve data for each class and returns a display that can be plotted with
CalibrationDisplay.plot().- Parameters:
- data_source{“test”, “train”}, default=”test”
The data source for the calibration curve.
- n_binsint or “auto”, default=”auto”
The number of bins to use for the calibration curve.
"auto"derives the number of bins from the data source size using the cube-root rule.- strategy{“uniform”, “quantile”}, default=”quantile”
The strategy to use for binning.
- Returns:
CalibrationDisplayThe calibration curve display.
CalibrationDisplay.plot()for more details.