An analyst for an insurance company’s data science team is developing a model to predict whether a submitted claim is fraudulent. Generally, less than 10% of claims are fraudulent. The threshold value for classifying an observation as fraudulent is not universally agreed upon. Which of the following is a good way to evaluate a model for classifying claims as fraudulent or not?

a. Precision

b. Specificity

c. Accuracy

d. Area under the ROC curve