After threshold tuning, F1 changes but ROC-AUC stays identical. Is that consistent?
Classic MLLevel: IntermediateThreshold versus ranking
After threshold tuning, F1 changes but ROC-AUC stays identical. Is that consistent?
Hints
- Separate scores from class labels.
- Which metric consumes each output?
Answer
Yes. Threshold tuning changes hard labels, not the underlying scores or their ranking. ROC-AUC evaluates scores across thresholds, while F1 evaluates one decision rule.
Source and licence: scikit-learn 1.7.2 — classification threshold · BSD-3-Clause · answer: Quantoros editorial