K-IPO: Kendall-constrained Importance Preserving Oversampling for Imbalanced Tabular Data

2026-07-17

AI Summary

This work addresses the degradation of original feature importance rankings caused by conventional oversampling methods on imbalanced tabular data. To preserve interpretability, the authors propose a generate-then-filter oversampling framework that explicitly constrains the Kendall’s tau correlation between the feature importance rankings of synthetic and original samples, with stronger protection afforded to highly important features. The framework adopts a generator-agnostic “generate-and-filter” strategy, making it compatible with diverse model interpretation techniques. Experimental results across 20 datasets demonstrate that the proposed method significantly improves fidelity in feature importance preservation, enhances explanation consistency, and boosts class separability, all while achieving better predictive performance without incurring prohibitive computational overhead.

Abstract

Oversampling is widely used to address class imbalance in tabular classification, but existing methods can distort the feature importance ranking underlying model explanations. Although recent studies have quantified this distortion by comparing real and synthetic data, none have actively sought to prevent it. In this paper, we introduce Kendall-constrained Importance-Preserving Oversampling (K-IPO), a generator-agnostic, "generate-then-select" framework that preserves the original data's feature importance ranking during augmentation. K-IPO iteratively generates minority-class candidates and accepts them only if their inclusion maintains a user-defined minimum Kendall's tau (τ) correlation with the reference ranking. Optionally, stricter constraints can be applied to the highest-ranked features. We evaluated K-IPO on 20 imbalanced binary classification datasets using three classifiers and multiple explanation methods. In most cases, K-IPO achieved the best or tied-best results in feature importance preservation, explanation consistency, and class separability. It also generally improved predictive performance while maintaining competitive computational overhead.

Problem

• class imbalance
• oversampling
• feature importance
• Kendall's tau
• tabular data

Innovation

• Kendall's tau
• feature importance preservation
• oversampling
• imbalanced tabular data
• explanation consistency