imbalanced-learn

Why we included this project

Most classifiers quietly learn to predict the majority class when one class dwarfs the other, and imbalanced-learn exists to counter that. It sits inside the scikit-learn ecosystem, so teams who already know sklearn's fit/predict pattern can add resampling and ensemble methods like SMOTE, ADASYN, and random under- or over-sampling without learning a new framework. The API mirrors sklearn closely, which makes it easy to drop into existing pipelines and cross-validation loops. For fraud detection, medical diagnostics, churn prediction, or any problem where the rare case is the one that matters, the user guide and example gallery help you pick a sensible strategy. It is a focused, mature library rather than a sprawling framework, which suits teams whose bottleneck is class imbalance, not model architecture.