Definition
The ROC curve shows how a classifier's sensitivity (share of correctly caught positives) and false-positive rate change as the decision threshold moves. The area under the curve (AUC) summarizes model quality in one number: 0.5 corresponds to chance, 1 to perfect discrimination; values above 0.8 are considered good.
Data mining & machine learning
The advantage of AUC is independence from the chosen threshold and from the class ratio, which overall accuracy lacks — with 2 % positives the model "everything negative" is 98 % accurate. You then pick the threshold by the cost of errors: for disease screening you favour sensitivity, for loan approval specificity.
Complementary tools are the confusion matrix, the gain/lift curve used in marketing and the precision-recall curve for very imbalanced classes. Compare models on test data, not training data.
In Statistica
ROC curves with AUC and classification matrices are provided by logistic regression, trees, boosting, random forests and neural networks; the Data Mining menu also has a Rapid Deployment and model comparison node that draws ROC, gain and lift for several models in one chart and computes the optimal threshold.
Related terms
- Logistic regressionLogistic regression models the probability of a binary outcome — a patient responds to treatment or not, a customer…
- Cross-validationCross-validation estimates how well a model will perform on new data.
- Decision treesA decision tree splits data by a sequence of simple rules ("age > 45 and income < 30 thousand") into ever more…
- Random forests and boostingRandom forests and boosting are ensemble methods that combine hundreds of decision trees into one more accurate model.
Knowledgebase guides
FAQ
- What AUC is sufficient?
- It depends on the task: 0.7 may be usable in marketing, while medicine and credit risk expect 0.8 or more. Always compare with current practice.
- How do I choose the threshold?
- By the cost ratio of false positives to false negatives; the Youden index maximizes the sum of sensitivity and specificity.
Try it on your own data
Statistica free for 30 days
Full version, no credit card. Or get a pricing estimate in a minute.
Data mining & machine learning
Updated: September 2026.