Databricks Machine Learning Associate Question 166
Single answerYou are working with a classification model in Databricks and need to evaluate its performance. The business team has set a designated threshold of 0.6 for predicting the 'positive' class. After evaluating the model, you notice the following metrics at that threshold: precision = 0.75, recall = 0.68, and F1-score = 0.71. The team emphasizes minimizing false negatives over false positives. What steps should you take to adjust the threshold appropriately?
- A
Lower the threshold to increase recall at the cost of precision.
- B
Raise the threshold to increase precision at the cost of recall.
- C
Keep the threshold at 0.6, as it provides a balance between precision and recall.
- D
Switch to a regression model instead of a classification model.
Show answer and explanation
Correct answer: A
Explanation
To minimize false negatives, recall must be prioritized over precision. Lowering the threshold increases recall by allowing more predictions to be classified as 'positive,' thereby reducing false negatives. Raising or maintaining the threshold would not fulfill the business requirement. Switching to a regression model is irrelevant in this context.
- A. Correct.
Lowering the threshold will classify more instances as 'positive,' increasing recall and reducing false negatives. This aligns with the business goal of minimizing false negatives.
- B. Incorrect.
Raising the threshold will increase precision but reduce recall, which would lead to more false negatives. This contradicts the business requirement of minimizing false negatives.
- C. Incorrect.
Keeping the threshold at 0.6 maintains the current metrics, but it does not address the business's emphasis on reducing false negatives. Therefore, this is not the optimal choice.
- D. Incorrect.
Switching to a regression model is unrelated to the problem, as the task is classification. This is not a viable solution.