Closed casper-hansen closed 5 years ago
Support for predict_proba
on model must be implemented to get a y_score.
Support added for binary roc_auc_score
in dev
Support added for multiclass scoring metrics. Use 'multiclass_average' in cv_options and specify either ‘micro’, ‘macro’, ‘samples’ or ‘weighted’
Go through each metric from sklearn.metrics and make sure every metric works.
Supported metrics: the ones with
metric(y_true, y_pred)
(i.e. roc_auc_score doesn't work, since it implements a
metric(y_true, y_score)
)