cp-based-fair-maker 0.0.4

A library that provides metrics and algorithms for bias assessment in datasets using CP-based methods

cp-based-fair-maker

Description

This is a library for creating fair ML datasets using Constraint Programming (CP). It suggests a way to create fair datasets by solving a constraint satisfaction problem (CSP) using CP and various fairness metrics. The library is based on the Google OR-Tools CP solver.

Glossary

1. Protected attribute: An attribute/feature that divides the population into groups that we want to be fair to. For example, race or gender can be protected attributes.
2. Privileged group: Considering a protected attribute, the groups that have historically been advantaged or have had more opportunities.
3. Unprivileged group: Considering a protected attribute, the groups that have historically been disadvantaged or have had fewer opportunities.
4. Instance: A row in the dataset.
5. Feature: A column in the dataset.
6. Instance weight: A weight assigned to each instance in the dataset. It can be seen as the number of times an instance is repeated in the dataset.
7. Label column: The column that represents the class labels of the instances.
8. Score columns: Columns that represent the confidence scores of the instances for each class label.
9. Fairness metric: A metric that measures the fairness of the dataset.
10. Fairness threshold: A threshold value for a fairness metric. The fairness metric value is compared to this threshold to determine if the dataset is fair or not.

Install from source

If you want to install the package from the source code, you can clone the repository and install it using pip:

pip install build setuptools wheel
python -m build
pip install dist/cp_based_fair_maker-<version>.whl

Here <version> should be replaced with the actual version number of the package.