Choosing a Primary Model Evaluation Metric for Business Impact: Accuracy vs F1 vs ROC-AUC

06 Eyl 08:00

I'm setting up a new predictive model for a fintech use case where both false positives and false negatives have cost implications. When deciding the primary evaluation metric, which of the following would you prioritize? A) Accuracy – overall correct predictions. B) F1 Score – balance between precision and recall. C) ROC‑AUC – ability to discriminate across thresholds. Please share which metric you would choose and why it fits the business context best.