AUDITING AI SYSTEMS

Dataset

COMPASS RACIAL BIAS PREDICTION WITH MACHINE LEARNING

Overview

This repository contains code and analysis for auditing AI systems, specifically focusing on the Compass dataset for predicting racial bias with machine learning models. The goal is to assess the fairness and potential biases present in machine learning algorithms trained on this dataset.

Structure

• Dataset: This directory holds the dataset files. Ensure that the necessary dataset files are placed in this directory before running the code.
• notebooks: Jupyter notebooks for interactive analysis and visualization
• The source code directory contains Python scripts that perform data preprocessing, model training, fairness evaluation, and bias analysis. The main components include:
• Feature_preprocessing.ipynb
• Data_and_task_setup.ipynb: Code for cleaning and preparing the dataset.
• model_Selection.ipynb: Implementation of machine learning models and training process.
• requirements.txt: Lists the required Python packages and their versions for reproducibility.