Balancing of an imbalanced dataset by applying SMOTE variants and predicting neonatal mortality using ensemble learning techniques

Dynamic environment and imbalanced datasets are unavoidable challenges in developing medical diagnostic tools where incremental learning is a necessity. The prediction tools upon imbalanced data normally work with majority class bias, and it is not easy to recognize faulty classes. This work aims to solve the class imbalance problem by generating synthetic data using SMOTE variants to balance the dataset and predict the neonatal mortality by adopting different ensemble classification methods. This system will be applied to diagnose newborns, vulnerable to die in the initial period of 28 days after birth.