Brief

Researchers have developed a novel data augmentation technique called Binary Gaussian Copula Synthesis (BGCS) specifically for binary clinical data, aiming to improve early dialysis prediction in chronic kidney disease (CKD). This method addresses the class imbalance issue common in medical datasets by generating synthetic minority-class samples that preserve pairwise dependencies among binary features. A fine-tuned GPT-2 model then filters these synthetic samples for clinical plausibility, leading to improved predictive performance and distributional fidelity compared to existing methods. The enhanced model was integrated into a decision support system, highlighting key predictive features like electrolyte imbalances and cardiovascular comorbidities. AI