Fair-Scoring

Fairness metrics for continuous risk scores.

The implemented algorithms are described in the paper [1].

Project Links

Documentation | PyPI | Paper

Quickstart

Installation

Install with pip directly:

pip install fair-scoring

Example Usage

The following example shows how compute the equal opportunity bias of the compas dataset

import pandas as pd
from fairscoring.metrics import bias_metric_eo

# Load compas data
dataURL = 'https://raw.githubusercontent.com/propublica/compas-analysis/master/compas-scores-two-years.csv'
df = pd.read_csv(dataURL)

# Relevant data
scores = df['decile_score']
target = df['two_year_recid']
attribute = df['race']

# Compute the bias
bias = bias_metric_eo(scores, target, attribute, groups=['African-American', 'Caucasian'], favorable_target=0,
                      prefer_high_scores=False)

Further examples

Further examples - especially the experiments conducted for the publication - can be found in the documentation.

Development

Setup

Clone the repository and install from this source via

pip install -e .[dev]

Tests

To execute the tests install the package in development mode (see above)

pytest

Following the pytest framework, tests for each package are located in a subpackages named test

Docs

To build the docs move to the ./docs subfolder and call

make clean
make html

References

[1] Becker, A.-K. and Dumitrasc, O. and Broelemann, K.; Standardized Interpretable Fairness Measures for Continuous Risk Scores; Proceedings of the 41th International Conference on Machine Learning, 2024; pdf

@inproceedings{Becker2024FairScoring,
    author = {Ann{-}Kristin Becker and Oana Dumitrasc and Klaus Broelemann}
    title  = {Standardized Interpretable Fairness Measures for Continuous Risk Scores},
    booktitle={Proceedings of the 41th International Conference on Machine Learning},
    year = {2024}
}