Weighted F1 is high while macro F1 is poor. What pattern should you investigate?
Classic MLLevel: IntermediateMinority-sensitive averaging
Weighted F1 is high while macro F1 is poor. What pattern should you investigate?
Hints
- Compare the weights assigned to classes.
- Aggregate scores can conceal a failing class.
Answer
Support-weighted averaging can be dominated by frequent classes; macro gives each class equal weight. Inspect per-class support and scores for weak minority classes.
Source and licence: scikit-learn 1.7.2 — model evaluation · BSD-3-Clause · answer: Quantoros editorial