Overlapping, Rare Examples and Class Decomposition in Learning Classifiers from Imbalanced Data

Abstract

This paper deals with inducing classifiers from imbalanced data, where one class (a minority class) is under-represented in comparison to the remaining classes (majority classes). The minority class is usually of primary interest and it is required to recognize its members as accurately as possible. Class imbalance constitutes a difficulty for most algorithms learning classifiers as they are biased toward the majority classes. The first part of this study is devoted to discussing main properties of data that cause this difficulty. Following the review of earlier, related research several types of artificial, imbalanced data sets affected by critical factors have been generated. The decision trees and rule based classifiers have been generated from these data sets. Results of first experiments show that too small number of examples from the minority class is not the main source of difficulties. These results confirm the initial hypothesis saying the degradation of classification performance is more related to the minority class decomposition into small sub-parts. Another critical factor concerns presence of a relatively large number of borderline examples from the minority class in the overlapping region between classes, in particular for non-linear decision boundaries. The novel observation is showing the impact of rare examples from the minority class located inside the majority class. The experiments make visible that stepwise increasing the number of borderline and rare examples in the minority class has larger influence on the considered classifiers than increasing the decomposition of this class. The second part of this paper is devoted to studying an improvement of classifiers by pre-processing of such data with resampling methods. Next experiments examine the influence of the identified critical data factors on performance of 4 different pre-processing re-sampling methods: two versions of random over-sampling, focused under-sampling NCR and the hybrid method SPIDER. Results show that if data is sufficiently disturbed by borderline and rare examples SPIDER and partly NCR work better than over-sampling.

Multi-class and feature selection extensions of Roughly Balanced Bagging for imbalanced data

“…The degradation of classification performance is linked to other factors related to data distribution, such as the decomposition of the minority class into many rare sub-concepts playing a role of small disjuncts (Jo and Japkowicz 2004), the effect of too strong overlapping between the classes (Garcia et al 2007) or a presence of too many minority examples inside the majority class regions (Napierala and Stefanowski 2012). It has been shown that when these data difficulty factors occur together with class imbalance, they seriously hinder the recognition of the minority class Napierala et al 2010;Napierala and Stefanowski 2012;Stefanowski 2013Stefanowski , 2016a. In the experimental analysis of Roughly Balanced Bagging (see Section 3) we will refer to some difficulty factors by analysing types of unsafe examples in the distribution of the minority class following the methodology presented in Stefanowski (2012, 2016).…”

EPRENNID: An evolutionary prototype reduction based ensemble for nearest neighbor classification of imbalanced data

“…Different kinds of minority class examples may have a different influence on learning classifiers [62,63]. To enrich the performed analysis, we have further divided the datasets into three subgroups, dense, medium and sparse, which represent different degrees of difficulty to recognize minority elements.…”

“…Inspired by [62], we use the local neighborhood of minority elements to consider them as safe, borderline, rare or outliers. In this work, we propose an alternative definition that is more conservative than the one used in [62].…”

“…In this work, we propose an alternative definition that is more conservative than the one used in [62]. For each minority instance, we determine its five nearest neighbors in the dataset and denote it as:…”

Associative learning on imbalanced environments: An empirical study

“…Over the last years, however, research on this topic has also put the emphasis on studying the effect of imbalance together with other data complexity characteristics such as overlapping, small disjuncts and noisy data (He et al, 2015;López et al, 2013;Napierala et al, 2010;Prati et al, 2004;Stefanowski, 2013). Another critical subject that has attracted increasing interest in the scientific community is how to assess the performance of a classification model in the presence of imbalanced data sets because most common metrics (e.g., accuracy and error rates) strongly depend on the class distribution and assume equal misclassification costs, which may lead to distorted conclusions (He and Garcia, 2009;Menardi and Torelli, 2014).…”