Hybrid Machine Learning Models for Accurate Type 2 Diabetes Mellitus Prediction Using a Stacking Classifier and a Meta-Model Approach

Abstract

Type 2 diabetes mellitus (T2DM) is a major global health issue that significantly reduces life expectancy and impairs overall quality of life. Early diagnosis of T2DM is critical, as it can help prevent or delay the development of associated complications. This study aims to assess the effectiveness of a stacked ensemble machine learning model, incorporating a meta-model approach, for the early detection of T2DM. In the first stage of modeling (Level 1), two base models-random forest (RF) and support vector machines (SVM)-are trained using distinct feature selection, tuning, and optimization strategies. In the second stage, ensemble methods, such as voting and stacking classifiers, are employed to combine the predictions from these base models. The final prediction is made by a meta-model, specifically a gradient boosting classifier (GBC), which classifies individuals as either positive or negative for T2DM. Further classification of positive instances is based on low-, moderate-, or high-risk levels. The system is designed for deployment to enable real-time assessment of diabetes risk. The Meta-Model (GBC), with an accuracy of 99.13%, a precision of 100%, specificity of 100%, and an F1 score of 99.25%, consistently outperforms the base models (SVM and RF) and ensemble models (voting and stacking classifiers), proving highly effective in the early detection of T2DM. This significant accuracy highlights the enormous promise of machine learning methods for assisting in the early identification of T2DM. The findings of this study will significantly enhance precision medicine screening for T2DM, empowering healthcare professionals to diagnose the condition early and ultimately improve patient outcomes.