ok3ks / bias_and_ml

This repository creates a machine learning model and further assesses the bias of this model with regards to the features

Bias_and_ML

This repository creates a machine learning model and further assesses the bias of this model with regards to the features. When viewing Bias and ML.ipynb, scroll to almost the end, red errors are due to environment not being uploaded.

GOAL

The goal of this project is to build a machine learning model which is not only acccurate and generalises well but also not biased to sensitve features in a dataset.

PROCESS In order to do this Algorithmic fairness methods such as reweighing,fairness without awareness was employed. in addition to state of the art machine learning algorithm XGBoost. XGBoost Model was manually tuned while assessing the impact of model parameters on model performance while

RESULT Results were obtained with and without algorithmic fairness methods and compared for insight Tradeoffs between generalisation, and fairness were also explored.