Evaluation

The evaluation support metrics for classification, regression, tasks. You can use our default set of metrics or use specified metrics.

Default Metric Set

We support following default metric set for binary, multi-class classification, regression tasks:

• Binary Classification
• AUC
• KS
• Confusion Matrix
• Gain
• Lift
• Precision Table
• Recall Table
• Accuracy Table

• FScore Table

• Multi Classification

• Accuracy
• Precision

• Recall

• Regression

• RMSE
• MAE
• MSE
• R2Score

Specify them in the 'default_eval_setting' parameter: 'binary', 'regression', 'multi'

Metrics

You can also set metrics you want to use in the 'metrics' parameter. These metrics are available:

auc
multi_accuracy
multi_recall
multi_precision
binary_accuracy
binary_recall
binary_precision
multi_f1_score
binary_f1_score
ks
confusion_matrix
lift
gain
biclass_precision_table
biclass_recall_table
biclass_accuracy_table
fscore_table
rmse
mse
mae
r2_score