GARP Risk and AI (RAI) Practice Exam 2026 - Free RAI Practice Questions and Study Guide

In binary classification, a correctly identified positive instance is called a true positive. It means the model predicted the positive class and the item truly belongs to that positive class, capturing the event of interest accurately. This concept is foundational for metrics like precision and recall: precision measures how many of the predicted positives are truly positive, while recall measures how many of the actual positives were captured. For example, in fraud detection, a true positive would be a fraudulent transaction that is correctly flagged as fraud. The other outcomes would be a false positive (flagged as fraud but actually legitimate), a true negative (correctly identified as legitimate), and a false negative (fraudulent but not flagged).