Class Imbalances versus Class Overlapping: An Analysis of a Learning System Behavior

By Ronaldo Cristiano Prati, Gustavo E. A. P. A. Batista, Maria Carolina Monard. 2004

356 citations recorded by the source corpus.

Subfields: Accounting, Artificial Intelligence

How researchers see this paper

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

In Accounting

In Artificial Intelligence

Citations in the corpus

Box drawings for learning with imbalanced data (2014, background): “Many works discuss problems caused by class imbalance [22, 16].”
DBSMOTE: Density-Based Synthetic Minority Over-sampling TEchnique (2012, background): “Over-sampling along the shortest path avoids the over-lapping problem [28] because this shortest path is the skeleton path [1].”
Analysis of preprocessing vs. cost-sensitive learning for imbalanced classification. Open problems on intrinsic data characteristics (2012, background): “…several investigations also suggest that there are other factors that contribute to such performance degradation, for example, size of the dataset, class imbalance level, small disjuncts, density, and overlap complexity (Japkowicz & Stephen, 2002; Prati & Batista, 2004; Weiss & Provost, 2003).”
A study of the behavior of several methods for balancing machine learning training data (2004, background): “In [18] we developed a systematic study aiming to question whether class imbalances hinder classiﬁer induction or whether these deﬁciencies might be explained in other ways.”
Handling imbalanced datasets: A review (2006, background): “Prati et al [25] developed a systematic study aiming to question whether class imbalances hinder classifier induction or whether these deficiencies might be explained in other ways.”
Cost-sensitive learning methods for imbalanced data (2010, background): “These problems have also been reported by [9], [33], [40].”
Safe-Level-SMOTE: Safe-Level-Synthetic Minority Over-Sampling TEchnique for Handling the Class Imbalanced Problem (2009, methodology): “In this paper, we consider only two-class case [5], [ 17 ].”
Feature selection for high-dimensional class-imbalanced data sets using Support Vector Machines (2014, background): “This problem is especially challenging when both classes have a high degree of overlap as has been pointed out e.g. by Prati et al. [25] and Qu et al. [26].”
Data mining with imbalanced class distributions: concepts and methods (2009, background): “The argument against a direct relationship between class imbalance and classifiers performance degradation is treated in [7], where a broad empirical study is carried out to question whether class imbalances are a problem by themselves and are truly to blame for the loss of performance of learning a”
An insight into classification with imbalanced data: Empirical results and current trends on using data intrinsic characteristics (2013, background): “Particularly, in [102] one can find a study where the authors propose several experiments with synthetic datasets varying the imbalance ratio and the overlap existing between the two classes.”
Exploring the effects of quantity and location of pseudo-absences and sampling biases on the performance of distribution models with limited point occurrence data (2011, background): “Machine learning algorithms perform well in very unbalanced designs (Prati et al. 2004), and an appropriate threshold to convert derived probabilities into a binomial variable can be used to avoid inaccuracies (Jiménez-Valverde & Lobo 2006).”
To Combat Multi-Class Imbalanced Problems by Means of Over-Sampling Techniques (2016, background): “Also, it generates synthetic samples regardless of the majority class; consequently, the over-lapping between different classes increase remarkably [41].”
Computer Aided Diagnostic Support System for Skin Cancer: A Review of Techniques and Algorithms (2013, background): “Several studies have demonstrated that the accuracy degradation on unbalanced data sets is more severe when the classes overlap significantly [190, 194, 195] which is the case in skin lesion classification.”
Learning from Imbalanced Data (2009, background): “The existence of within-class imbalances is closely intertwined with the problem of small disjuncts, which has been shown to greatly depreciate classification performance [23], [27], [28], [29].”
Cost-sensitive boosting for classification of imbalanced data (2007, methodology): “For decision trees, the class distributions at each leaf is usually used as the score [34].”