Answer-first summary for fast verification
Answer: AUC_weighted
For imbalanced classification problems like credit card fraud detection, accuracy (option C) is misleading as it can be high even if the model fails to detect the minority class (fraud cases). AUC_weighted (option B) is optimal because it calculates the area under the ROC curve for each class and weights them by the number of true instances, making it robust to class imbalance. This aligns with Azure ML best practices for automated ML, where AUC_weighted is recommended for imbalanced datasets. The community discussion confirms this, with high upvotes for comments supporting AUC_weighted and noting its robustness. Options A, D, and E are regression metrics and unsuitable for classification tasks.
Author: LeetQuiz Editorial Team
Ultimate access to all questions.
You are creating a classification model to identify credit card fraud for a banking company using automated machine learning in Azure Machine Learning. The training dataset is highly imbalanced.
Which primary metric should you use to evaluate the model?
A
normalized_mean_absolute_error
B
AUC_weighted
C
accuracy
D
normalized_root_mean_squared_error
E
spearman_correlation
No comments yet.