Adaptive Synthetic Sampling with Generative Adversarial Networks (AS-GAN) for Predicting Chronic Kidney Disease on Unbalanced Data

A major issue in medical data analysis is class imbalance, particularly when identifying chronic kidney disease (CKD). With a disproportionately high percentage of cases falling into the majority class, class labels are commonly imbalanced in medical datasets. The impact of class imbalance on training data is examined in this study, which also offers a method for creating a neural network classifier for CKD medical decision-making. To address this challenge, we focus on combining the Generative Adversarial Network with the Adaptive Synthetic Sampling Technique (ASS). To enhance ASS, we introduce Adaptive Synthetic-GAN, a novel two-phase oversampling method that leverages GAN. When there is insufficient minority-class data for GAN to handle alone, the GAN component converts the unrealistic samples produced by ASS into a more realistic distribution of data. Performance measures including accuracy, precision, recall, and f1 score of 0.98, 0.99, 0.98, and 0.99 were attained by the suggested model AS-GAN.