scikit-learn-contrib/imbalanced-learn
Imbalanced Learn
imbalanced-learn is a dataset balancing framework and Python machine learning extension designed to resample training data and reduce the impact of class imbalance. It provides a toolkit of algorithms for adjusting class distributions to improve model performance on minority class prediction.
As a scikit-learn resampling library, it extends the ecosystem with specialized tools for balancing datasets through over-sampling and under-sampling techniques. This allows for the correction of skewed class proportions to reduce model bias toward the majority class.
The library implements the scikit-learn estimator API with fit and transform methods and provides pipeline-compatible transformers. Its capabilities cover data preprocessing and machine learning resampling, including randomized oversampling logic and heuristic undersampling mechanisms.
Features
- Dataset Class Balancing - Provides tools to adjust class distributions in machine learning data to improve minority class prediction.
- Class Imbalance Handling - Provides a programmatic framework to resample training data and reduce the impact of class imbalance.
- Class Ratio Scaling - Adjusts the number of samples per class to achieve a target distribution before training a classifier.
- Machine Learning Resampling - Applies oversampling or undersampling techniques to create a balanced training set for scikit-learn models.
- Undersampling Heuristics - Implements heuristic undersampling mechanisms to reduce majority class samples based on density or distance.
- Python Machine Learning Libraries - Extends the scikit-learn ecosystem with advanced sampling methods to mitigate dataset bias.
- Imbalanced Data Sampling - Applies sampling techniques to datasets with significant class differences to increase model accuracy.
- Oversampling Logic - Generates synthetic minority samples by interpolating between existing data points to expand the minority class boundary.
- Machine Learning Model APIs - Implements standard fit and transform methods for compatibility with the scikit-learn estimator ecosystem.
- Minority Class Prediction - Improves the ability of a classifier to accurately detect rare events or underrepresented categories in data.
- Data Transformation Pipelines - Wraps sampling logic into objects that can be integrated into sequential data processing workflows.
- Imbalanced Datasets - Module for under-sampling and over-sampling techniques.
- Data Validation and Cleaning - Resampling techniques for imbalanced machine learning datasets.
- Feature Engineering Tools - Handling imbalanced datasets.
