Databricks Machine Learning Associate Question 252
Select 2You are working on a binary classification problem to detect fraudulent transactions. The dataset is highly imbalanced, with only 2% of transactions labeled as fraudulent. Which of the following evaluation metrics are most appropriate for assessing the performance of your model in this scenario?
- A
Accuracy, as it provides a clear measure of the proportion of correct predictions.
- B
Recall, as it measures the model's ability to identify fraudulent transactions.
- C
Precision, as it measures the proportion of correctly identified fraudulent transactions out of all predicted fraudulent transactions.
- D
F1 Score, as it balances the trade-off between recall and precision.
- E
R-squared, as it explains the proportion of variance in the fraudulent transaction data.
Show answer and explanation
Correct answers: B, D
Explanation
In imbalanced classification problems like fraud detection, recall is essential to ensure that fraudulent transactions are detected, while the F1 Score helps to balance recall and precision, especially when false negatives are more critical than false positives. Accuracy and R-squared are not suitable metrics for this scenario.
- A. Incorrect.
Accuracy is not suitable for imbalanced datasets, as a model predicting all transactions as non-fraudulent would achieve high accuracy but fail to identify fraudulent transactions.
- B. Correct.
Recall is crucial in this case, as the primary objective is to minimize false negatives (i.e., fraudulent transactions that go undetected).
- C. Incorrect.
Precision is important but not as critical as recall in this scenario, as the main focus is on detecting as many fraudulent transactions as possible, even at the expense of some false positives.
- D. Correct.
F1 Score is appropriate because it accounts for both precision and recall, providing a balanced evaluation metric for imbalanced datasets.
- E. Incorrect.
R-squared is a regression metric and is not applicable to binary classification problems.