Which evaluation metric does Databricks AutoML optimize by default for a binary classification problem?
Select an answer to reveal the explanation.
Short Explanation and Infographic
AutoML picks ROC AUC as its default classification metric because it's threshold-independent and handles class imbalance better than accuracy — smart defaults matter when you're not tuning manually.
Full explanation below image
Full Explanation
Databricks AutoML uses ROC AUC as the default metric for binary classification because: 1) Threshold-independence — AUC measures the model's ability to rank positive examples above negatives at all thresholds, not just at 0.5. 2) Class imbalance robustness — unlike accuracy, AUC isn't dominated by the majority class. 3) Widely accepted — AUC is the standard evaluation metric in most binary classification settings. AutoML supports other metrics via the primary_metric parameter: 'f1', 'log_loss', 'precision', 'accuracy', 'roc_auc'. For regression, AutoML uses R² by default. For multiclass, it uses F1 (macro). The user can override the default: from databricks import automl; summary = automl.classify(df, target_col='label', primary_metric='f1', timeout_minutes=30). Accuracy is not the default due to imbalanced dataset issues. Log loss is a valid but secondary choice.