Enhanced Wasserstein Generative Adversarial Network (EWGAN) to Oversample Imbalanced Datasets

This paper examines WGAN as a more advanced technique for addressing imbalanced data sets in the context of machine learning. A variety of domains, including medical diagnosis and image generation, are affected by the problem of imbalanced datasets since it is essential to represent the minority class to train a satisfactory model and create various types of data. To overcome these challenges WGAN uses some features such as; Residual connections in the critic network, better sampling for minority classes, and some noise and sample reshaping. These innovations contribute to the increased stability of the model, the quality of synthetic data, and the distribution of classes in a dataset. The comparative analysis of WGAN with basic GAN and Improved GAN has shown the effectiveness of the given algorithm in terms of producing high-quality diversified synthetic data that is closer to the real data distribution. The study identifies the future research direction of WGAN in enhancing machine learning based on reliable and diverse synthesized data, providing new insights and directions for future studies and practical applications in tackling data imbalance issues.