bluecast.conformal_prediction.conformal_prediction

Module Contents

Classes

ConformalPredictionWrapper

Calibrate a model instance given a calibration set.
class bluecast.conformal_prediction.conformal_prediction.ConformalPredictionWrapper(model: Any, nonconformity_measure_scorer: Callable = hinge_loss, random_seed: int = 20)

Bases: bluecast.conformal_prediction.base_classes.ConformalPredictionWrapperBaseClass

Calibrate a model instance given a calibration set.

Parameters:

  • model – An already fitted model instance of any type

  • nonconformity_measure_scorer – A function object to calculate nonconformity scores with args y_calibration, preds

plot_non_conformity_scores(nonconformity_scores: List[float]) → None

Plot the distribution of nonconformity scores. :param nonconformity_scores: List of nonconformity scores

calibrate(x_calibration: pandas.DataFrame, y_calibration: pandas.Series)

Calibrate a model instance given a calibration set.

Parameters:

  • x_calibration – Calibration set features. Must be unseen data for the model

  • y_calibration – Calibration set labels or values

predict(x)
predict_proba(x)
predict_interval(x: pandas.DataFrame) → numpy.ndarray
predict_sets(x: pandas.DataFrame, alpha: float = 0.05) → numpy.ndarray