A SYSTEM FOR MITIGATING DISPARATE IMPACT IN ALGORITHMIC DECISIONS

The present disclosure relates to a system for mitigating disparate impact in algorithmic decision. The aim of the present invention is to solve the problem of combinatorial discrepancy in algorithms. The system uses linear programming tools to mitigate disparate impact, while also considering weighted version of the problem. The proposed system is implemented on real world dataset to assess its performance and perform evaluation, wherein the results showed that the proposed system provides solution to the problems such like fairness gerrymandering problem (to a certain extent) and data size blow-up for multiple polyvalent sensitive attributes can be solved without compromising on the accuracy of the final classification results