A company wants to catch as many fraudulent transactions as possible with its ML model. Which evaluation metric should be prioritized to maximize identified fraud cases?

Recall (also known as sensitivity or true positive rate) is the most appropriate metric when the goal is to identify as many positive cases (fraudulent transactions) as possible. It measures the proportion of actual positive cases that were correctly identified by the model (True Positives / (True Positives + False Negatives)). Maximizing recall minimizes the number of missed fraudulent transactions (False Negatives). F1 score is the harmonic mean of precision and recall, providing a balanced view, but it wouldn't prioritize catching all fraud. AUC measures the model's ability to distinguish between classes across various thresholds, which is a good overall performance indicator but doesn't specifically prioritize minimizing false negatives. Precision measures the proportion of positive identifications that were actually correct (True Positives / (True Positives + False Positives)), which would be important if minimizing false alarms was the priority, but not for maximizing fraud detection.