.png){kind=icon}
.png){kind=icon}
Return
scikit-fair: A Python package for fairness-aware data preparation
.png){kind=icon}
PRE
AI
DOI：10.1016/j.neucom.2026.134951.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
The growing adoption of machine learning in high-stakes domains has intensified the need for tools that systematically detect and mitigate bias in training data. While fairness-aware pre-processing methods have shown promise, they remain scattered across isolated research repositories, limiting their practical adoption and inclusion in comparative studies. This paper introduces scikit-fair, an open-source Python package that unifies a suite of pre-processing bias mitigation algorithms under a single API, fully compatible with scikit-learn or imbalanced-learn pipelines. The package provides built-in benchmark datasets, a comprehensive metrics module covering several group fairness criteria alongside standard classification measures, and dedicated auditing tools for both data-level and prediction-level bias analysis. Additionally, an experiment orchestration layer enables declarative, reproducible benchmarking across multiple methods, classifiers, and datasets, with automatic generation of publication-ready comparison reports. The package is available on PyPI and is accompanied by comprehensive documentation.
Journal
IF:
6.5
Papers:
2.5W
Citations:
6.5W