Area under the ROC Curve (ROC-AUC)
The short answer · 16 words
In short
Evaluation indicator for how well a classification model separates positive and negative cases across all thresholds.
Flow-alerta.cab (2026): Area under the ROC Curve (ROC-AUC). https://flow-alerta.cab/en/glossar/area-under-the-roc-curve-roc-auc. Editorially reviewed on: 09 October 2026.
How much detail would you like?
Practical
High for comparing classification models, but it does not replace a suitable decision threshold, nor the testing of costs, calibration and class distribution.
Example scenario: A company compares two models for detecting customers at risk of migration on the same test dataset and adds precision, recall and a cost matrix to ROC-AUC.
Model comparison for fraud, quality, lead or failure forecasts. For operation, additional threshold values, wrong decision costs and time stability must be checked.
At a glance
- Core
- ROC-AUC is the area under the receiver operating characteristic curve. It summarizes how well predictive values rank positive cases higher than negative ones. AUC alone is more general and must always be named with the curve used.
- Use
- High for comparing classification models, but it does not replace a suitable decision threshold, nor the testing of costs, calibration and class distribution.
- Example
- Example scenario: A company compares two models for detecting customers at risk of migration on the same test dataset and adds precision, recall and a cost matrix to ROC-AUC.
- Limits
- Name AUC without curve type; confuse ROC-AUC with PR-AUC or Average Precision; ignore unbalanced classes; compare models on different test data; derive good calibration from a high value.
Questions about this term
Is ROC-AUC equal to Accuracy?
No.
Is it always higher than better?
Only with the same task and clean test design.
What is important for very rare positive cases?
Additionally, consider Precision-Recall metrics.
All details and background
ROC-AUC is the area under the receiver operating characteristic curve. It summarizes how well predictive values rank positive cases higher than negative ones. AUC alone is more general and must always be named with the curve used.
ELI5 – Explained simply
The model ranks cases according to a predictive value. ROC-AUC summarizes how well it separates positive and negative cases across different decision thresholds.
Why it matters in practice
High for comparing classification models, but it does not replace a suitable decision threshold, nor the testing of costs, calibration and class distribution.
Example
Example scenario: A company compares two models for detecting customers at risk of migration on the same test dataset and adds precision, recall and a cost matrix to ROC-AUC.
Pitfalls and limits
- Name AUC without curve type; confuse ROC-AUC with PR-AUC or Average Precision; ignore unbalanced classes; compare models on different test data; derive good calibration from a high value.
Origin & history
ROC evaluations describe the ratio of hit and false alarm rates when the decision threshold is changed. In libraries such as scikit-learn, ROC-AUC is implemented as a code for classification models.
Use in mid-sized companies
- Model comparison for fraud, quality, lead or failure forecasts. For operation, additional threshold values, wrong decision costs and time stability must be checked.
Related terms
- ROC curve
- Precision
- Recall
- Calibration
- Confusion matrix
Matching Flow-alerta service
AI opportunity analysis & roadmapAI consulting: AI potential analysis & roadmap on tasks, data and prerequisites of a possible first AI use case.Matching tool / use case
View use casesTool: scikit-learn; Use Case: comparison of two binary classifiers on an unchanged test datasetSources
- scikit-learn.orgscikit-learn, roc_auc_score
- scikit-learn.orgscikit-learn, Modellbewertung
Quellenprüfung: 09.10.2026. Definition und dokumentierte Einordnung abgeglichen; Beispiele beschreiben mögliche Anwendungen.
Editorially reviewed on: