Databricks Machine Learning Associate Question 606
Single answerYou are building a model to predict customer churn for a subscription-based service. The goal is to minimize false negatives, as failing to identify customers likely to churn could result in significant revenue loss. Which evaluation metric would be the most appropriate to optimize in this scenario?
- A
Accuracy
- B
Precision
- C
Recall
- D
F1-Score
Show answer and explanation
Correct answer: C
Explanation
Since the objective is to minimize false negatives (i.e., ensure that most customers likely to churn are identified), recall is the most appropriate metric. Recall focuses on capturing as many true positives as possible, which is crucial in this scenario to prevent revenue loss.
- A. Incorrect.
Accuracy measures the proportion of correctly classified instances out of all instances, but it does not specifically focus on minimizing false negatives. In an imbalanced dataset, accuracy might be misleading.
- B. Incorrect.
Precision measures the proportion of true positives out of all predicted positives. While it is important in some contexts, it does not address the concern of minimizing false negatives.
- C. Correct.
Recall measures the proportion of true positives out of all actual positives. Optimizing recall directly helps in minimizing false negatives, which aligns with the goal of reducing missed churn predictions.
- D. Incorrect.
The F1-Score balances precision and recall, but if the primary objective is to minimize false negatives, recall alone is a more direct and appropriate metric.