Monitoring Realized Performance for Binary Classification

We currently support the following standard metrics for bianry classification performance calculation:

  • roc_auc

  • f1

  • precision

  • recall

  • specificity

  • accuracy

For more information about estimating these metrics, refer to the Calculating Standard Performance Metrics for Binary Classification section.

We also support the following complex metrics for binary classification performance calculation:

  • confusion_matrix: a metric that has four components (TP, FP, FN, TN).

  • business_value: a metric that combines the four components of the confusion matrix using user-specified weights for each element, allowing for a connection between model performance and business results.

For more information about estimating these metrics, refer to the Calculating Confusion Matrix Elements for Binary Classification and Calculating Business Value for Binary Classification sections.