DataTerminology.comNEW
Also known as: Area Under the ROC Curve, AUC-ROC
A scalar summary of the ROC curve representing classifier discrimination across all possible thresholds.
DESCRIPTION
ROC AUC measures a binary classifier’s ability to rank positive instances higher than negatives, independent of a specific threshold. It integrates true positive rate versus false positive rate over all thresholds. AUC of 0.5 indicates random guessing; 1.0 indicates perfect discrimination. ROC AUC is robust to class imbalance for ranking but can be misleading under skewed costs. Complement with PR AUC and calibration metrics for comprehensive evaluation.
ALTERNATIVE DEFINITIONS
Wikipedia
The area under the ROC curve (AUC) provides an aggregate measure of performance across all classification thresholds.
Highlights threshold-agnostic aggregate performance.
Google Developers
AUC is the area under the ROC curve; it measures the entire two-dimensional area underneath the entire ROC curve (from (0,0) to (1,1)).
Emphasizes geometric interpretation of AUC as area.
RELATED TERMS
ROC Curve Recall Specificity Confusion Matrix Threshold Binary Classification Class Imbalance Calibration Discrimination
COMPETING TERMS
RELATIONS
- Broader: AUC
KNOWLEDGE AREAS
Data Science AI Mathematics Testing & Validation
TARGET AUDIENCE
Data Scientist AI Engineer MLOps Engineer BI Specialist Data Analyst
EXTERNAL SOURCES
- Wikipedia - Receiver operating characteristic
- scikit-learn - ROC and AUC metric documentation
- Google Developers - ML Crash Course on ROC/AUC
- Wikipedia - Background on ROC and AUC
TOP QUESTIONS
When should I prefer ROC AUC over PR AUC?
Use ROC AUC when class distribution is relatively balanced and you care about ranking across all thresholds; PR AUC is better under heavy class imbalance.
Category: Comparison
What does an ROC AUC of 0.5 mean?
It indicates random ranking performance with no discriminative ability.
Category: Clarification
How does ROC AUC relate to ranking quality?
ROC AUC equals the probability a random positive scores higher than a random negative.
Category: Clarification
What are common pitfalls when using ROC AUC?
Relying solely on AUC can hide poor precision at relevant thresholds; combine with PR AUC and domain-specific metrics.
Category: Mistakes
How many thresholds does ROC AUC consider?
It integrates over all possible thresholds, making it threshold-independent.
Category: Clarification
Does calibration affect ROC AUC?
Calibration doesn’t change ranking; it may not affect AUC but impacts decision thresholds and interpretability.
Category: Opinion
How to compute ROC AUC robustly?
Use cross-validation, stratification, and avoid data leakage; plot ROC curves for multiple folds.
Category: HowTo
Can ROC AUC be misleading?
Yes, for highly imbalanced data it may look high while positive precision is low; check PR AUC.
Category: Risk
What is micro vs macro ROC AUC?
Extensions aggregate across classes differently: micro aggregates predictions, macro averages per-class AUCs.
Category: Clarification