PROJECT TITLE :

Imbalanced Data Classification via Cooperative Interaction Between Classifier and Generator

ABSTRACT:

When learning classifiers from imbalanced data, there is a greater likelihood of a significant bias toward the majority class. Several solutions based on generative adversarial networks have been suggested as a potential solution to this problem (GANs). The currently available GAN-based methods, on the other hand, do not make efficient use of the connection that exists between a classifier and a generator. In this article, a novel three-player structure called a discriminator, a generator, and a classifier, along with decision boundary regularization, are proposed as potential solutions to current problems. Our approach is unique in that the generator is trained along with the classifier in order to provide minority samples that gradually expand the minority decision region, which improves performance for the classification of imbalanced data. The performance of the proposed method is superior to that of the existing methods when applied to both real data sets and synthetic imbalanced data sets.