Metric Support¶
Responsible AI Tracker reports the following metrics depending on the prediction problem:
- Classification: accuracy, precision, recall, F1, roc_auc, logloss
For binary classification, the averaging parameter in precision, recall, and F1 scores in sklearn is set to “binary” (i.e., average = “binary”). For the roc_auc score the multiclass parameter is set to “raise” (i.e., multi_class = “raise”).
For multiclass classification, the averaging parameter in precision, recall, and F1 scores in sklearn is set to “macro” (i.e., average = “macro”). For the roc_auc score the multiclass parameter is set to “ovr” (i.e., multi_class = “ovr”).
Regression: mean-squared error (mse), root mean-squared error (rmse), mean absolute error (mae), r-squared (r2)