INTRODUCTION - Imbalanced Learning: Foundations, Algorithms, and Applications - page 4

particular data example [21, 22]. Research in the past indicates that there is
a strong connection between cost-sensitive learning and imbalanced learning
[4, 23, 24]. In general, there are three categories of approaches to implement
cost-sensitive learning for imbalanced data. The first class of techniques applies
misclassification costs to the dataset as a form of dataspace weighting ( trans-
lation theorem [25]); these techniques are essentially cost-sensitive bootstrap
sampling approaches where misclassification costs are used to select the best
training distribution. The second class applies cost-minimizing techniques to the
combination schemes of ensemble methods ( Metacost framework [26]); this class
consists of various meta techniques, such as the AdaC1, AdaC2, and AdaC3
methods [27] and AdaCost [28]. The third class of techniques incorporates cost-
sensitive functions or features directly into classification paradigms to essentially
“fit” the cost-sensitive framework into these classifiers, such as the cost-sensitive
decision trees [21, 24], cost-sensitive neural networks [29, 30], cost-sensitive
Bayesian classifiers [31, 32], and cost-sensitive support vector machines (SVMs)
[33-35].
1.2.3 Kernel-Based Learning Methods
There have been many studies that integrate kernel-based learning methods
with general sampling and ensemble techniques for imbalanced learning.
Some examples include the SMOTE with different costs (SDC) method [36]
and the ensembles of over/undersampled SVMs [37, 38]. For example, the
SDC algorithm uses different error costs [36] for different classes to bias the
SVM to guarantee a more well-defined boundary. The granular support vector
machines—repetitive undersampling (GSVM-RU) algorithm was proposed in
[39] to integrate SVM learning with undersampling methods. Another major
category of kernel-based learning research efforts focuses more concretely on
the mechanisms of the SVM itself; this group of methods are often called
kernel modification methods , such as the kernel classifier construction algorithm
proposed in [40]. Other examples of kernel modification include the various
techniques used for adjusting the SVM class boundary [41, 42]. Furthermore,
the total margin-based adaptive fuzzy SVM (TAF-SVM) kernel method was
proposed in [43] to improve SVM robustness. Other major kernel modification
methods include the k -category proximal SVM (PSVM) [44], SVMs for extreme
imbalanced datasets [45], support cluster machines (SCMs) [46], kernel neural
gas (KNG) algorithm [47], hybrid kernel machine ensemble (HKME) algorithm
[48], and the Adaboost relevance vector machine (RVM) [49].
1.2.4 Active Learning Methods
Active learning methods have also been proposed for imbalanced learning in
the literature [50-53]. For instance, Ertekin et al. [51, 52] proposed an efficient
SVM-based active learning method that queries a small pool of data at each