skfp.metrics.auroc_score#

skfp.metrics.auroc_score(y_true: ndarray, y_score: ndarray, *args, constant_target_behavior: str | float = nan, **kwargs) → float#

Area Under Receiver Operating Characteristic curve (AUROC / ROC AUC).

Wrapper around scikit-learn roc_auc_score function, which adds constant_target_behavior to control behavior for all-zero y_true labels. Scikit-learn behavior is to throw an error, since AUROC is undefined there, but this can easily happen for cross-validation in imbalanced problems.

Parameters:

y_true (array-like of shape (n_samples,) or (n_samples, n_outputs)) – Ground truth (correct) target values.
y_score (array-like of shape (n_samples,) or (n_samples, n_outputs)) – Target scores, i.e. probability of the class with the greater label for each output** of the classifier.
*args – Any additional parameters for the underlying roc_auc_score function.
**kwargs – Any additional parameters for the underlying roc_auc_score function.
constant_target_behavior ("raise", np.NaN, None, or float, default=np.NaN) – Value returned if y_true contains only constant values. “raise” is the default scikit-learn behavior, which raises an error. Default np.NaN (or None) ignores the given fold in cross-validation. Alternatively, float value (e.g. 0.5, 1.0) can be returned.

Returns:

score – AUROC value.

Return type:

float

Examples

>>> import numpy as np
>>> from skfp.metrics import auroc_score
>>> y_true = np.array([0, 0, 0])
>>> y_score = np.array([0.5, 0.6, 0.7])
>>> auroc_score(y_true, y_score)
nan
>>> auroc_score(y_true, y_score, constant_target_behavior=0.5)
0.5