FairML: A Julia Package for Fair Classification

TL;DR

FairML.jl is a Julia package that offers a comprehensive framework for fair classification, addressing data imbalance, model fairness during training, and post-processing adjustments to mitigate unfairness in machine learning predictions.

Contribution

The paper introduces FairML.jl, a novel Julia package that integrates preprocessing, in-processing, and post-processing stages for fair classification, enhancing fairness control in ML workflows.

Findings

Resampling method effectively reduces data imbalance unfairness.

Incorporating fair ML methods improves model fairness during training.

Combining phases yields better fairness-performance trade-offs.

Abstract

In this paper, we propose FairML.jl, a Julia package providing a framework for fair classification in machine learning. In this framework, the fair learning process is divided into three stages. Each stage aims to reduce unfairness, such as disparate impact and disparate mistreatment, in the final prediction. For the preprocessing stage, we present a resampling method that addresses unfairness coming from data imbalances. The in-processing phase consist of a classification method. This can be either one coming from the MLJ.jl package, or a user defined one. For this phase, we incorporate fair ML methods that can handle unfairness to a certain degree through their optimization process. In the post-processing, we discuss the choice of the cut-off value for fair prediction. With simulations, we show the performance of the single phases and their combinations.