The goal of this project is to attempt to consolidate fairness related metrics, transformers and models into a package that (hopefully) will become a contribution project to scikit-learn.
Fairness, in data science, is a complex unsolved problem for which many tactics are proposed - each with their own advantage and disadvantages. This packages aims to make these tactics readily available, therefore enabling users to try and evaluate different fairness techniques.
The documentation for this project can be found here.
Consider all the steps in a machine learning pipeline.
This package will offer tools at every step to make the pipeline more fair.
We have datasets available that will help you benchmark your fairness tools.
skfair.datasets.load_arrests
We have filtering techniques that try to filter out information that correlates with sensitive attributes.
skfair.preprocessing.InformationFilter
We have models you're able to constrain with regards to a fairness metric.
skfair.linear_model.DemographicParityClassifier
skfair.linear_model.EqualOpportunityClassifier
We have meta estimators that allow you to correct the model after it has been trained.
We offer metrics that are designed to measure unfairness in your dataset.
skfair.metrics.equal_opportunity_score
skfair.metrics.p_percent_score