Generative Adversarial Neural Networks based Oversampling Technique for Imbalanced Credit Card Dataset

Abstract

The imbalanced dataset is a challenging issue in many classification tasks. Because it leads a machine learning algorithm to poor generalization and performance. The imbalanced dataset is characterized as having a huge difference between the number of samples that contain each class. Unfortunately, various resampling methods are proposed to solve this problem. In our work, we target enhancing the handling of the imbalanced dataset using a new oversampling technique based on generative adversarial neural networks. Our method is benchmarked against the widely used oversampling technique including the synthetic minority oversampling technique (SMOTE), random oversampling technique (ROS), and the adaptive synthetic sampling approach(ADSYN). Additionally, three machine learning algorithms are used for evaluation. The outcome of our experiments on a real-world credit card dataset shows the strong ability of the proposed solution against the competitive oversampling techniques to overcome the imbalanced problem in the European credit card dataset.