Abstract
Without any specific way for imbalance data classification, artificialintelligence algorithm cannot recognize data from minority classes easily. Ingeneral, modifying the existing algorithm by assuming that the training data isimbalanced, is the only way to handle imbalance data. However, for a normaldata handling, this way mostly produces a deficient result. In this research,we propose a class expert generative adversarial network (CE-GAN) as thesolution for imbalance data classification. CE-GAN is a modification in deeplearning algorithm architecture that does not have an assumption that thetraining data is imbalance data. Moreover, CE-GAN is designed to identify moredetail about the character of each class before classification step. CE-GAN hasbeen proved in this research to give a good performance for imbalance dataclassification.