Databricks Machine Learning Associate Question 253
Select 2You are tasked with evaluating a machine learning model used to classify fraudulent transactions. The dataset is highly imbalanced, with fraudulent transactions being a small minority. Which of the following statements correctly describe the use of Recall and F1 as evaluation metrics in this scenario?
- A
Recall is important because it measures how many actual fraudulent transactions were correctly identified by the model.
- B
F1 combines Recall and Precision, providing a balanced evaluation metric for imbalanced datasets.
- C
Recall alone is sufficient to evaluate the performance of the model in this scenario.
- D
F1 is not suitable for imbalanced datasets as it gives equal importance to Recall and Precision.
- E
A high F1 score indicates that the model performs well in both identifying true positives and avoiding false positives.
Show answer and explanation
Correct answers: A, B
Explanation
In scenarios with imbalanced datasets, like fraud detection, Recall is critical to ensure that as many true positive cases as possible are identified. However, Recall alone is not sufficient because it does not consider Precision, which measures the accuracy of positive predictions. F1 combines Recall and Precision into a single metric, making it suitable for evaluating models in such scenarios. This ensures a balanced view of the model's performance, particularly in cases where both missing true positives and generating false positives have significant consequences.
- A. Correct.
Correct. Recall is crucial in this scenario as it measures the model's ability to detect actual fraudulent transactions, which is important in imbalanced datasets where missing fraud cases can have severe consequences.
- B. Correct.
Correct. F1 is a harmonic mean of Recall and Precision, making it a useful metric to balance the trade-off between these two measures, especially in imbalanced datasets.
- C. Incorrect.
Incorrect. Recall alone does not provide a complete picture of the model's performance, as it does not account for Precision (how many predicted positives are actually correct).
- D. Incorrect.
Incorrect. F1 is specifically designed to handle imbalanced datasets by balancing Recall and Precision, making it suitable for this type of problem.
- E. Incorrect.
Incorrect. While a high F1 score reflects good performance in balancing Recall and Precision, it does not directly indicate avoidance of false positives; Precision measures that aspect.