Universum based kernelized weighted extreme learning machine for imbalanced datasets

Imbalanced classification is a challenging problem in the fields of machine learning and data mining. Cost-sensitive methods can handle this issue by considering different misclassification costs of classes. Various modifications of support vector machine (SVM) and extreme learning machine (ELM) have been proposed to handle the class imbalance problem, which focuses on different aspects to resolve the class imbalance. Such as Weighted ELM (WELM) and weighted SVM (WSVM). The Universum SVM (USVM) incorporates the prior information in the classification model by adding Universum sample to the training sample to handle the class imbalance problem. Various other modifications of SVM have been proposed, which use Universum sample in the classification model generation. Moreover, the existing ELM-based classification models intended to tackle class imbalance do not consider the prior information about the sample distribution for training. An ELM-based classification model creates two symmetry planes, one for each class. The Universum-based ELM classification model tries to create a third plane between the two symmetric planes using Universum sample. This paper proposes a novel hybrid framework called Universum-based kernelized WELM (UKWELM) and Universum-based reduced kernelized WELM (URKWELM), which combines the Universum learning with WELM for the first time to inherit the advantages of both techniques. Universum samples are the training samples from the same domain but they do not belong to any of the target classes. The proposed UKWELM, URKWELM, and other classifiers in consideration are evaluated by using Knowledge Extraction based on Evolutionary Learning dataset repository. The experimental results demonstrate that UKWELM and URKWELM achieve better performance in contrast to the rest of the classifiers for imbalance learning.