Home▸Machine Learning & Neural Networks▸Classifier Threshold Sweep

Model Evaluation Simulator

Sweep a decision threshold over a classifier's real score distribution and watch the ROC curve and precision-recall curve build point by point, with live AUC and average precision computed by trapezoidal integration.

Machine Learning & Neural Networks2DEasy60 FPS📱 Mobile-adapted⇄ 3D version

Evaluating a classifier means more than reading off one accuracy number — it means understanding how every possible decision threshold trades false positives against false negatives. This simulator generates a realistic pair of score distributions (one per class), then sweeps a threshold across them to build the ROC curve and the precision-recall curve point by point, computing AUC-ROC and average precision by real trapezoidal integration rather than a lookup table. Drag the threshold to see the confusion matrix update live, or let it auto-sweep to watch both curves trace out from nothing. Adjust class separation, score noise and the positive-class prior to see exactly when ROC and PR agree — and when they don't.

⚙ Under the hood

Sweep a classifier's decision threshold over its score distribution to generate ROC and precision-recall curves with trapezoidal AUC integration.

model-evaluationclassificationROC curve

2D · HTML5 Canvas 2D · 60 FPS target · runs fully client-side, no install