Analysis Of Fairness Measures

Read Paper

Reference

The project's synthetic dataset was generated using the code from this repository. Hence, the generator is not present here.

The goal

1. be able to choose the right measure of fairness depending on various conditions

2. eventually, answer whether the data is balanced (or how much it is probable to be fair)

Set-up

• After cloning the repository, create a virtual environment inside: python3.11 -m venv venv

• Don't forget to activate it (depending on your OS) e.g. source venv/bin/activate

• Install the dependencies: pip install -r requirements.txt

File-structure

Remark: related to n = 24 dataset will be referenced to as sample, while n = 56, which we worked with, will be main.

• utils script contains (mainly metric) functions

• metricCalculation is how we got the calculations

• calculations directory contains sample metric calculations

• plotting is code for getting plots out of calculations

• plots directory contains 2 sub-directories:

  1. n24 are sample plots

  2. n56 are main plots

• data directory contains the sample dataset itself

• resourseConsumption is experiments on time and memory complexity for generating datasets