ROC Curve Explorer

KNN · Bias–Variance Tradeoff · Decision Boundary

Trials: 0

Binary Classification

Controls

K Neighbors 5

K=1 (overfit)K=51 (underfit)

Training Set Size 150

30500

Current AUC

—

Test Accuracy

—

AUC Mean

—

AUC Std Dev

—

Feature Space & Decision Boundary

K = 5

Class 0 train

Class 1 train

Class 0 test

Class 1 test

Decision boundary

ROC Curve

FPR vs TPR

Current model

Trial curves

Random chance

Bias–Variance Low K → jagged boundary, high variance (ROC curves spread wide across trials). High K → smooth boundary, high bias. Watch the boundary morph as you drag K.

Reading the ROC Each point is a classification threshold. AUC summarises discriminability — 1.0 is perfect, 0.5 is random chance (dashed diagonal).

100 Trials Resampling training data while holding the test set fixed reveals how much the ROC varies — a direct measure of model variance for the chosen K.