Skip to content

check_class_proba_prediction_bounds

yohou.testing.class_proba.check_class_proba_prediction_bounds(forecaster, y_test, X_actual_test=None)

Check all probability values are in [0, 1].

Parameters

Name Type Description Default
forecaster BaseClassProbaForecaster

Fitted class-probability forecaster instance.

 required
y_test DataFrame

Test target data.

 required
X_actual_test DataFrame or None

Test features.

 None

Raises

Type Description
AssertionError

If any probability value is outside [0, 1].

Source Code

View on GitHub

Show/Hide source 
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
def check_class_proba_prediction_bounds(
    forecaster, y_test: pl.DataFrame, X_actual_test: pl.DataFrame | None = None
) -> None:
    """Check all probability values are in [0, 1].

    Parameters
    ----------
    forecaster : BaseClassProbaForecaster
        Fitted class-probability forecaster instance.
    y_test : pl.DataFrame
        Test target data.
    X_actual_test : pl.DataFrame or None, default=None
        Test features.

    Raises
    ------
    AssertionError
        If any probability value is outside [0, 1].

    """
    forecasting_horizon = min(3, len(y_test))
    y_pred = forecaster.predict_class_proba(forecasting_horizon=forecasting_horizon)

    proba_cols = [col for col in y_pred.columns if "_proba_" in col]
    assert len(proba_cols) > 0, "No probability columns found"

    for col in proba_cols:
        values = y_pred[col]
        min_val = values.min()
        max_val = values.max()
        assert min_val >= 0.0, f"Probability column {col} has negative values (min={min_val})"
        assert max_val <= 1.0, f"Probability column {col} has values > 1 (max={max_val})"