Feature Selection Under Fairness Constraints

Learning from large dimensional data presents major challenges related to the size of the data. Thus, dimensionality reduction techniques such as feature selection are brought in to reduce computation time, improve prediction performance, and better understand the data. However, two problems can occur with current feature selection methods when protected features are considered. The presence of protected features among the selected ones which often lead to unfair results and the presence of redundant features which carry potentially the same information with the protected ones. By protected features we mean features on which it is important to not have bias due to data imbalances. In view of such issues, we introduce a fair feature selection method that takes into account the existence of protected features and their redundant. Our new method finds a set of relevant features with no protected features with the least possible redundancy under prediction quality constraint. This constraint consists of a trade-off between fairness and prediction performance. Below, we outline the strategy of our trade-off approach.