Which metrics should the data scientist use to optimize the model?

Explanations:

Specificity measures the ability of the model to identify true negatives, which is not the priority when trying to capture as many positives as possible.

The false positive rate is a measure of how often negatives are incorrectly classified as positives, which is less relevant when the focus is on detecting positives.

Accuracy is not ideal for imbalanced classes, as it may be high even if the model fails to capture most of the fraudulent transactions (positives).

The area under the precision-recall curve is a good metric for evaluating models in imbalanced datasets, as it focuses on the model’s performance in detecting positives.

The true positive rate (also known as recall or sensitivity) measures the proportion of actual positives that are correctly identified, which aligns with the goal of capturing as many fraudulent transactions as possible.