Addressing Imbalance in Multi-Label Data via Label-Specific Distance-based Oversampling
Researchers have developed a new oversampling technique called Label-Specific Distance-based Multi-Label Oversampling (LSDMLO) to address the challenge of imbalanced data in multi-label classification. Traditional methods often overlook the varying importance of features for different labels, leading to confusion and overfitting. LSDMLO aims to create more accurate synthetic instances by considering label-specific distances and weighted feature spaces, thereby improving the performance of multi-label classifiers. AI
IMPACT Improves multi-label classification accuracy by addressing data imbalance.