What is precision?

What is precision?

• A. The proportion of predicted positives that are correct.
• B. The proportion of actual positives that are predicted correctly.
• C. The proportion of predicted negatives that are correct.
• D. The proportion of all predictions that are positive.

Answer: (A)

Precision measures how exact the positive predictions are. It answers the question: when the model predicts something as positive, what fraction actually is positive? It’s calculated as true positives divided by all predicted positives (TP / (TP + FP)). This focuses on the cost of false positives, so you want a high precision when incorrectly labeling negatives as positives is costly. For example, in spam filtering, high precision means most emails marked as spam really are spam, minimizing false alarms. This differs from recall, which asks: of all actual positives, how many did we identify? Recall = TP / (TP + FN). The other options describe unrelated ideas: the proportion of actual positives predicted correctly is recall, not precision; the proportion of predicted negatives that are correct concerns negative predictions, not positives; and the proportion of all predictions that are positive is just how often the model predicts positive, not how accurate those positives are.

Precision measures how exact the positive predictions are. It answers the question: when the model predicts something as positive, what fraction actually is positive? It’s calculated as true positives divided by all predicted positives (TP / (TP + FP)). This focuses on the cost of false positives, so you want a high precision when incorrectly labeling negatives as positives is costly. For example, in spam filtering, high precision means most emails marked as spam really are spam, minimizing false alarms.

This differs from recall, which asks: of all actual positives, how many did we identify? Recall = TP / (TP + FN). The other options describe unrelated ideas: the proportion of actual positives predicted correctly is recall, not precision; the proportion of predicted negatives that are correct concerns negative predictions, not positives; and the proportion of all predictions that are positive is just how often the model predicts positive, not how accurate those positives are.