Metrics

fast_automl.metrics.check_scoring

def fast_automl.metrics.check_scoring(scoring, classifier) [source]

Creates a default regression or classifier scoring rule. This is R-squared for regressors and ROC AUC for classifiers.

Parameters:

scoring : str or callable

A str (see scikit-learn's model evaluation docs) or a scorer callable with signature scorer(estimator, X, y) which returns a single value.

classifier : bool

Indicates that the estimator is a classifier.

def fast_automl.metrics.roc_auc_score(y, output) [source]

Parameters:

y : array-like of shape (n_samples,)

target values.

output : array-like of shape (n_samples, n_classes)

Predicted probability of each class.

Returns:

score : scalar

ROC AUC score, default is one-versus-rest for multi-class problems.