Imbalanced Classification through the Lens of Spurious Correlations
This addresses the challenge of imbalanced classification for machine learning practitioners, offering a novel perspective that is largely overlooked but incremental in its approach.
The paper tackles the problem of unreliable classification performance due to class imbalance by viewing it as a condition that amplifies Clever Hans effects, and proposes an Explainable AI-based method to identify and eliminate these effects, achieving competitive performance on three datasets.
Class imbalance poses a fundamental challenge in machine learning, frequently leading to unreliable classification performance. While prior methods focus on data- or loss-reweighting schemes, we view imbalance as a data condition that amplifies Clever Hans (CH) effects by underspecification of minority classes. In a counterfactual explanations-based approach, we propose to leverage Explainable AI to jointly identify and eliminate CH effects emerging under imbalance. Our method achieves competitive classification performance on three datasets and demonstrates how CH effects emerge under imbalance, a perspective largely overlooked by existing approaches.