Close to Reality: Interpretable and Feasible Data Augmentation for Imbalanced Learning

Fuente: arXiv

Gespeichert in:

Bibliographische Detailangaben
Hauptverfasser:	da Silva, Matheus Camilo, Costanzo, Gabriel Gustavo, de Lorenzo, Andrea, Junior, Sylvio Barbon
Format:	Preprint
Veröffentlicht:	2026
Schlagworte:	Machine Learning
Online-Zugang:	Acceder al recurso
Tags:	Tag hinzufügen Keine Tags, Fügen Sie den ersten Tag hinzu!

_version_	1866912966152028160
author	da Silva, Matheus Camilo Costanzo, Gabriel Gustavo de Lorenzo, Andrea Junior, Sylvio Barbon
author_facet	da Silva, Matheus Camilo Costanzo, Gabriel Gustavo de Lorenzo, Andrea Junior, Sylvio Barbon
contents	Many machine learning classification tasks involve imbalanced datasets, which are often subject to over-sampling techniques aimed at improving model performance. However, these techniques are prone to generating unrealistic or infeasible samples. Furthermore, they often function as black boxes, lacking interpretability in their procedures. This opacity makes it difficult to track their effectiveness and provide necessary adjustments, and they may ultimately fail to yield significant performance improvements. To bridge this gap, we introduce the Decision Predicate Graphs for Data Augmentation (DPG-da), a framework that extracts interpretable decision predicates from trained models to capture domain rules and enforce them during sample generation. This design ensures that over-sampled data remain diverse, constraint-satisfying, and interpretable. In experiments on synthetic and real-world benchmark datasets, DPG-da consistently improves classification performance over traditional over-sampling methods, while guaranteeing logical validity and offering clear, interpretable explanations of the over-sampled data.
format	Preprint
id	arxiv_https___arxiv_org_abs_2603_13927
institution	arXiv
publishDate	2026
record_format	arxiv
spellingShingle	Close to Reality: Interpretable and Feasible Data Augmentation for Imbalanced Learning da Silva, Matheus Camilo Costanzo, Gabriel Gustavo de Lorenzo, Andrea Junior, Sylvio Barbon Machine Learning Many machine learning classification tasks involve imbalanced datasets, which are often subject to over-sampling techniques aimed at improving model performance. However, these techniques are prone to generating unrealistic or infeasible samples. Furthermore, they often function as black boxes, lacking interpretability in their procedures. This opacity makes it difficult to track their effectiveness and provide necessary adjustments, and they may ultimately fail to yield significant performance improvements. To bridge this gap, we introduce the Decision Predicate Graphs for Data Augmentation (DPG-da), a framework that extracts interpretable decision predicates from trained models to capture domain rules and enforce them during sample generation. This design ensures that over-sampled data remain diverse, constraint-satisfying, and interpretable. In experiments on synthetic and real-world benchmark datasets, DPG-da consistently improves classification performance over traditional over-sampling methods, while guaranteeing logical validity and offering clear, interpretable explanations of the over-sampled data.
title	Close to Reality: Interpretable and Feasible Data Augmentation for Imbalanced Learning
topic	Machine Learning
url	https://arxiv.org/abs/2603.13927