Preprocessing Matters: Automated Pipeline Selection for Fair Classification

Vladimiro González-Zelaya, Julián Salas, Dennis Prangle, Paolo Missier

Abstract: Improving fairness by manipulating the preprocessing stages of classification pipelines is an active area of research, closely related to AutoML. We propose a genetic optimisation algorithm, FairPipes, which optimises for user-defined combinations of fairness and accuracy and for multiple definitions of fairness, providing flexibility in the fairness-accuracy trade-off. FairPipes heuristically searches through a large space of pipeline configurations, achieving near-optimality efficiently, presenting the user with an estimate of the solutions’ Pareto front. We also observe that the optimal pipelines differ for different datasets, suggesting that no “universal best” pipeline exists and confirming that FairPipes fills a niche in the fairness-aware AutoML space.

External IDs:dblp:conf/mdai/Gonzalez-Zelaya23