Additional Optimizing Metrics for Binary Classification: Precision or Recall

[judahr]

In binary classification, there are situations where the data are imbalanced and only false positives or only false negatives are costly, we would want to tune the model on simply Precision or Recall. This would better optimize models based on these scenarios.

In Advanced training options, under the optimizing metric include a Precision option and a Recall option. Use this to tune the Model Builder for higher precision or higher recall.

[LittleLittleCloud]

@JakeRadMSFT @luisquintanilla

[luisquintanilla]

Makes sense to me. @LittleLittleCloud is it just a matter of adding those options in the UI or is work required in the AutoML API?

[LittleLittleCloud]

Just adding options on UI