ROC & AUC
Core route · model intuition
Threshold sweeps and AUC as a single interactive flow.
This lab turns evaluation metrics into something you can drag. Move the classification threshold, watch the ROC curve trace out, and connect AUC back to the separation you see in the score distributions.
At a glance
Why this matters
ROC and AUC are often taught as formulas. This makes the curve feel like a direct consequence of score separation, threshold choice, and class balance.
What to try first
Sweep the threshold and watch the ROC point move. Then shift class balance or overlap to see how AUC reacts and why it can be misleading when used alone.
Time and level
About 8 to 12 minutes. Beginner friendly. It works well as a first metrics note because the threshold sweep makes the curve feel earned.
Embed
Reading path
- Open the live interactive: https://kohnnn.github.io/interactive-explanation/roc-auc/
- Continue through Precision & Recall when you want the companion threshold metric after the ROC view
- Continue via Interactive or Visual Notes