Séminaire

Date : 31 mars 2025 14:30 - Salle :Salle A102 Less-Discriminatory Alternative and Interpretable Models for Binary Classification Taufiquar KHAN, chair Dpt. Math. & Stat. - UNC Charlotte, Charlotte, NC, USA)

In this presentation, we will discuss the recently proposed novel less-discriminatory, fair, and interpretable machine learning formulation for binary classification. We will discuss this new formulation that utilizes elementary biobjective programming methods and is implemented to find an optimal solution which maximizes the accuracy of the model while minimizing the bias. The optimal solution is computed by balancing between the accuracy and bias instead of choosing the balance heuristically. Furthermore, we will present the results of our sensitivity analysis on the Adverse Impact Ratio (AIR) in terms of the fairness parameter for the proposed model. This is joint work with Dr. Andrew Pangia (UNC Charlotte), Dr. Agus Sudjianto (UNC Charlotte and formerly at Wells Fargo Bank), and Dr. Aijun Zhang (Wells Fargo Bank).

Website of the speaker: https://math.charlotte.edu/directory/taufiquar-khan/