Gradually Generative Adversarial Networks Method for Imbalanced Datasets

Imbalanced dataset can cause obstacles to classification and result in a decrease performance. There are several methods that be used deal the data imbalances, such as based on SMOTE Generative Adversarial Networks (GAN). These for overcoming oversampling so amount of minority increase it reach balance with majority data. In this research, selected is classified small imbalanced less than 200 records. The proposed method Gradually Network (GradGAN) model which aims handle imbalances gradually. stages GradGAN adding original gradually will create new datasets until created. Based algorithm flow described, multiplied by value variable has been determined repeatedly produce balanced test results from an accuracy 8,3% when compare without GradGAN.