Multiclass Imbalance Problems: Analysis and Potential Solutions

By Shuo Wang, Xin Yao. IEEE Transactions on Systems Man and Cybernetics Part B (Cybernetics), 2012

553 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Electrical and Electronic Engineering

View Paper: https://www.semanticscholar.org/paper/ab8445a0b2d0e06377c21ce45832d1a4ccd3116a

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Electrical and Electronic Engineering

Citations in the corpus

Consistent Multiclass Algorithms for Complex Performance Measures (2015, methodology): “The G-mean measure is used to evaluate both binary and multiclass classifiers in settings with class imbalance (Sun et al., 2006; Wang & Yao, 2012), and is given by ψ GM (C) = n i=1 C ii n j=1 C ij 1/n .”
Binary relevance for multi-label learning: an overview (2018, methodology): “Next, a tri-class classi ﬁ er g : X × { 0 , + 1 , + 2 } (cid:3)→ R can be induced from D tri jk by applying a multi-class imbalance learning algorithm M I [59–61], i.e., g M I ( D tri jk ).”
COVID-19 identification in chest X-ray images on flat and hierarchical classification scenarios (2020, methodology): “In this situation, it is common that the classiﬁcation algorithm increases the likelihood of the most frequent classes and reduces the likelihood of the least frequent classes [11].”
Analysing the classification of imbalanced data-sets with multiple classes: Binarization techniques and ad-hoc approaches (2013, methodology): “For more details regarding this approach please refer to [58].”
To Combat Multi-Class Imbalanced Problems by Means of Over-Sampling Techniques (2016, background): “A high G-mean guarantees that no class is ignored [52].”
Learning with Complex Loss Functions and Constraints (2018, methodology): “For example in many retrieval tasks, it is desirable to optimize the F1-measure [23, 19]; in classification tasks with class imbalance, a classifier is often evaluated using metrics such as the G-mean or the H-mean measure that emphasize equitable performance on all classes [30, 32, 22].”
SMOTE for Learning from Imbalanced Data: Progress and Challenges, Marking the 15-year Anniversary (2018, background): “…the minority class against the remaining ones (One-versus-all approach), there are some extensions specifically employed for tackling multi-class imbalanced classification problems (Wang & Yao, 2012): the work from Fernández-Navarro et al. (2011), Alejo et al. (2015) and Abdi and Hashemi (2016).”
Boosting methods for multi-class imbalanced data classification: an experimental review (2020, background): “In this case, for example, a class may be a minority one when compared to some other classes, but a majority of the rest of them [13].”
Data-Driven Evolutionary Optimization: An Overview and Case Studies (2019, background): “Incomplete [24], imbalanced [25], [26], and noisy [27], [28] data bring errors to function evaluations of data-driven EAs, which may mislead the search.”