A company trains a classifier to detect credit card fraud; about 2% of transactions are fraudulent. The business priority is to capture as many fraudulent transactions as possible. Which evaluation metrics should the data scientist optimize? (Choose two.)

For fraud detection, where fraudulent transactions are rare (2%) but critical to identify, the goal is to capture as many fraudulent transactions as possible. This directly aligns with maximizing the True Positive Rate (Recall), which measures the proportion of actual fraudulent transactions that were correctly identified. While maximizing recall, it's also important to consider the precision of the model to avoid an overwhelming number of false alarms. The F1 Score is the harmonic mean of precision and recall, providing a balanced measure that is particularly useful when dealing with imbalanced datasets like this one. Optimizing F1 score ensures a good balance between catching most fraud and not flagging too many legitimate transactions incorrectly. Specificity and False Positive Rate are less critical here because they focus on correctly identifying non-fraudulent transactions or the rate of incorrect fraud flags, respectively. While important, they are secondary to the primary goal of catching fraud. Accuracy is misleading for imbalanced datasets because a model that always predicts "not fraudulent" would achieve 98% accuracy but miss all fraudulent transactions.