New applications of an oversampling method based on generative adversarial networks
Addisson Salazar, Luis Vergara, Gonzalo Safont · 2020
This paper presents new applications of a novel oversampling method based on Generative Adversarial Networks (GAN). Two challenging applications are approached: classification of stages of a neuropsychological test (Barcelona test) using electroencephalographic (EEG) data from epileptic patients; and classification of sleep stages using electrocardiographic (ECG) data from apnea patients. The results of the GAN-based method are compared with the commonly used interpolation-based method synthetic minority oversampling technique (SMOTE). The results show the superiority of the GAN method over SMOTE to improve the probability of error of the two-class classification proposed applications.