Precision-Recall AUC
Also known as: PR AUC, PR Curve
The Precision-Recall Area Under the Curve (PR AUC) is the area under the curve formed by plotting recall on the x-axis and precision on the y-axis. It is particularly useful for evaluating classifiers on imbalanced datasets, where it is often more informative than ROC AUC.
Key highlights
- More informative than ROC AUC for imbalanced datasets; focuses on the positive class.
- Sensitive to classifier performance on the minority class.
- Directly plots precision vs. recall, making the trade-off explicit.
- Useful for threshold selection based on domain-specific precision/recall requirements.
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Use PR AUC for imbalanced binary classification problems where the minority positive class is important. It is particularly valuable in fraud detection, rare disease diagnosis, and any domain where false negatives are costly but negatives vastly outnumber positives. PR AUC is often more informative than ROC AUC on imbalanced data.
Strengths & limitations
- More informative than ROC AUC for imbalanced datasets; focuses on the positive class.
- Sensitive to classifier performance on the minority class.
- Directly plots precision vs. recall, making the trade-off explicit.
- Useful for threshold selection based on domain-specific precision/recall requirements.
- Baseline (random classifier) varies with class imbalance, making comparison across datasets challenging.
- Not as widely adopted as ROC AUC; some practitioners may be less familiar with interpretation.
- Requires probability outputs or confidence scores, not just class predictions.
- Can be less stable than ROC AUC on very small positive class sizes.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
When should I use PR AUC instead of ROC AUC?
Use PR AUC for imbalanced datasets where the minority positive class is the focus of interest. ROC AUC can be misleading because a classifier can achieve high ROC AUC by correctly identifying the abundant negative class, even if positive class performance is poor. PR AUC directly focuses on the positive class performance.
What is the baseline PR AUC for a random classifier?
For a random classifier, PR AUC equals the proportion of positive samples (prevalence). For example, if 5% of samples are positive, a random classifier's expected PR AUC is 0.05. This varies with class imbalance, making PR AUC values less comparable across datasets with different prevalences than ROC AUC.
How do I choose a threshold from the PR curve?
Examine the precision-recall curve and choose a threshold that matches your requirements. If you need high precision (few false positives), choose a higher threshold on the right side of the curve. If you need high recall (few false negatives), choose a lower threshold on the left side. Your choice depends on the relative cost of false positives and false negatives.
Sources
- 1.Davis, J., & Goadrich, M. (2006). The relationship between precision-recall and ROC curves. Proceedings of the 23rd International Conference on Machine Learning, 233-240.
- 2.Saito, T., & Rehmsmeier, M. (2015). The precision-recall plot is more informative than the ROC plot when evaluating binary classifiers on imbalanced datasets. PLoS ONE, 10(3), e0118432.
You have read it. What now?
Cite this page
ScholarGate. (2026, June 3). Precision-Recall AUC. ScholarGate. https://scholargate.app/model-evaluation/precision-recall-auc