Contents

MultiLabelClassificationMetrics.ThresholdSelectionStrategy.precision(_:minimumRecall:)

A confidence threshold strategy for a specific precision that has at least a minimum recall value.

Declaration

case precision(Float, minimumRecall: Float)

Discussion

This strategy selects a threshold for each label by searching for the specified precision value on the label’s precision-recall curve. At the precision, the recall must be greater than or equal to the minimum recall value, otherwise a NaN threshold for the corresponding label is returned.

Use this strategy to reduce the rate of false-positive predictions while constraining the false-negative predictions.

See Also

Selection strategies