gluonts.nursery.anomaly_detection.supervised_metrics.bounded_pr_auc module#

gluonts.nursery.anomaly_detection.supervised_metrics.bounded_pr_auc.bounded_pr_auc(precisions: numpy.array, recalls: numpy.array, lower_bound: float = 0) → float[source]#

Bounded PR AUC –> AUC when recall > lower_bound.

Parameters

  • precisions (np.array) – precisions of different thresholds

  • recalls (np.array) – recalls of different thresholds

  • lower_bound (float) – lower bound of recalls

Returns

bounded PR-AUC

Return type

float