ROC / AUC

ROC / AUC is a fast, fully offline app that evaluates a binary classifier's scores with an ROC curve and its AUC - plus a live threshold slider, the confusion matrix, and a score-distribution view.

The signature view is the ROC curve itself: the trade-off between catching more positives and triggering more false alarms, shaded by its AUC area, with the optimal operating point marked.

WHAT'S INSIDE

ROC
The ROC curve with its shaded AUC area and the optimal point marked against your chosen threshold, plus the score distributions for each class and a plain-language verdict.

Data
A threshold slider (with a Youden-optimal hint) and an editable (score, class) table. AUC, accuracy, TPR, FPR, and the full confusion matrix update live as you move it.

Scores
The score-distribution view - positives and negatives plotted as overlapping histograms, so you can see exactly where they blur together.

Concept
A plain-English walkthrough of the threshold trade-off, TPR/FPR/AUC, and Youden's J - with the rate and Youden formulas rendered in offline KaTeX.

Simulate
Knobs for cases per class, separation, and spread - generate scored cases and watch the curve bow toward the corner as separation grows.

Examples
Ready-made datasets - a strong classifier, a coin flip, a medical test, and an imbalanced set - that load with one tap.

WHY ROC / AUC

- 100% offline - your data never leaves your device
- No account, no sign-up, no clutter
- Clean, distinctive interface built for students, researchers, and anyone evaluating a classifier
- Full ROC sweep, trapezoidal AUC, Youden's J, and the confusion matrix at any operating point
- Input is never lost - the threshold and every row are saved automatically

Whether you're a student learning why AUC beats accuracy on imbalanced data, a researcher picking an operating threshold, or just curious how a classifier's scores separate two classes, ROC / AUC shows you the whole picture in one glance.