Investigating Gender and Age Variability in Diabetes Prediction: A Multi-Model Ensemble Learning Approach

Abstract

The study investigates the intricate influence of gender and age variability in individuals diagnosed with diabetes, aiming to gain a comprehensive understanding of the diverse impact and implications of this prevalent metabolic disorder. A real-world dataset, obtained from a renowned diabetologist and meticulously maintained by Dr. Reddys' Lab, serves as the foundation for rigorous analysis. Leveraging the capabilities of ensemble learning, an advanced technique that combines multiple models, the predictive model's efficiency is substantially enhanced, resulting in precise and reliable predictions of individuals' diabetic status. Addressing the challenge of diabetes prediction, a novel ensemble learning model was proposed. The model combines the strengths of three distinct algorithms: Random Forest, Extra Trees, and Multilayer Perceptron (MLP). The model's output comprises a ternary label categorizing individuals as "diabetic, non-diabetic, or pre-diabetic", while the accompanying prediction score quantifies the likelihood of individuals belonging to each respective category. The findings of this research expand the existing body of knowledge on diabetes prediction, underscoring the untapped potential of ensemble learning methodologies in augmenting accuracy and predictive performance for diabetic patients.