Available in: GBM, DRF, Deep Learning, GLM, GAM, AutoML, XGBoost, Isolation Forest
This option specifies the metric to consider when early stopping is specified (i.e., when
stopping_rounds > 0). For example, given the following options:
then the model will stop training after reaching three scoring events in a row in which a model’s missclassication value does not improve by 1e-3. These stopping options are used to increase performance by restricting the number of models that get built.
Available options for
stopping_metric include the following:
AUTO: This defaults to
deviance(mean residual deviance) for regression, and
anomaly_scorefor Isolation Forest.
anomaly_score(for Isolation Forest only)
AUC(area under the ROC curve)
AUCPR(area under the Precision-Recall curve)
custom(for custom metric functions where “less is better”. It is expected that the lower bound is 0.) Note that this is currently only supported in the Python client for GBM and DRF. More information available in Python example below and here.
custom_increasing(for custom metric functions where “more is better”.) Note that this is currently only supported in the Python client for GBM and DRF. More information available in Python example below and here.
stopping_rounds must be enabled for
stopping_tolerance to work.