A stacking ensemble approach for diabetes prediction

Abstract

Diabetes Mellitus (DM) is a metabolic disorder which persist for longer duration in the human body due to fluctuating levels of blood glucose. Continuous monitoring of glucose levels is essential, as persistent hyperglycaemia can lead to complications such as retinopathy, nephropathy, and neuropathy. Therefore, emerging Machine Learning (ML) and data analytics methods are crucial for the identification, and management of DM. This research aims to develop predictive models that enable early intervention in diabetes management. A well-known biomedical dataset, PIMA, is used to implement a stacking ensemble method to improve diabetes classification. The stacking ensemble combines multiple diverse base algorithms to harness their collective predictive capabilities. Through comprehensive evaluation, it is found that the stacking ensemble method outperforms individual models and other ensemble techniques across various performance metrics. A metric value greater than 90% is obtained for accuracy, precision and recall. The experimental results highlight the potential of the stacking ensemble method as an effective model for accurate and reliable diabetes classification in biomedical data analysis.