Managing Imbalanced Data Sets in Multi-label Problems: A Case Study with the SMOTE Algorithm

Abstract

Multi-label learning has been becoming an increasingly active area into the machine learning community since a wide variety of real world problems are naturally multi-labeled. However, it is not uncommon to find disparities among the number of samples of each class, which constitutes an additional challenge for the learning algorithm. Smote is an oversampling technique that has been successfully applied for balancing single-labeled data sets, but has not been used in multi-label frameworks so far. In this work, several strategies are proposed and compared in order to generate synthetic samples for balancing data sets in the training of multi-label algorithms. Results show that a correct selection of seed samples for oversampling improves the classification performance of multi-label algorithms. The uniform generation oversampling, provides an efficient methodology for a wide scope of real world problems.

MLSMOTE: Approaching imbalanced multilabel learning through synthetic instance generation

“…By contrast the analysis conducted in [13] relies in a more sophisticated oversampling algorithm such as SMOTE, but taking into account only one minority label. Thus, it ignores the intrinsic nature of MLDs.…”

“…In [13], the authors analyze different strategies aimed to apply the original SMOTE algorithm to MLDs. They do it by means of three ways for selecting the seed instances, i.e.…”

“…The new samples can be clones of existent ones, or be synthetically produced as in SMOTE (Synthetic Minority Over-sampling Technique) [15]. Multilabel oversampling algorithms based on the cloning approach have been proposed in [12,13], being demonstrated its capability to deliver an improvement in classification results. A synthetic approach to produce new samples in MLDs is still to be faced.…”

“…It is also present in multilabel classification (MLC), since labels are unevenly distributed in most MLDs. To deal with imbalance in MLC, methods based on algorithmic adaptations [6][7][8], the use of ensembles [9,10], and resampling techniques [11][12][13] have been proposed.…”

“…The EML[9], BR-IRUS[10], and SmoteUG[13] algorithms were carefully implemented following the descriptions given in their respective papers.F. Charte et al / Knowledge-Based Systems xxx (2015) xxx-xxx…”