Supervised Metrics#

class oats.scorer.supervised_scorer.SupervisedScorer(delay: Optional[int] = None)[source]#

Bases: Scorer

Scorer for traditional supervised metrics; only useful if actual anomalies are known

Parameters:: delay (int, optional) – For pattern anomalies, how much tolerance to give for prediction; e.g. delay of 10 means only those predictions in the first 10 time steps of a pattern anomalies are counted as success. None if no delay needed. Defaults to None.

property f1#: F-1 Score

property fnr#: False Negative Rate

property fpr#: False Positive Rate

property precision#: Precision

process(preds, labels)[source]#

property recall#: Recall

property tnr#: True Negative Rate

property tpr#: True Positive Rate