Abstract: Accurate identification and prediction of diabetes complications contribute to improved patient health. However, existing prediction models predominantly employ single-task learning (STL) ...