Atlas · GenAI 2026
Class Imbalance Handling
Designing sampling, weighting, threshold and evaluation strategies when target classes have substantially unequal representation or costs.
Also searchable as: Class-Imbalance Handling, Imbalanced Classification, Imbalanced Data Handling
conceptModel Selection & TuningAI consensus: 0/3
Prerequisites
- mediumModel Evaluation
Accuracy can mislead under imbalance, so evaluation and metric selection are essential.
Recommended reference
imbalanced-learn: User guide — https://imbalanced-learn.org/stable/user_guide.html
Reviewed sources
Primary and first-party material reviewed for this editorial summary. These citations are separate from the AI consensus score above.
- imbalanced-learn: User guide
Resampling methods, evaluation and practical pitfalls for imbalanced classification.
- imbalanced-learn: Over-sampling
Resampling strategies and their assumptions.
- scikit-learn: Compute class weight
Class weighting as an alternative intervention to resampling.
Notes from AI deep research
Related skills
- ← is subcategory of: SMOTE
- → is subcategory of: Classification