An Ensemble Learning based Framework for Accurate Diabetic Diagnosis

Abstract

This paper presents a novel ensemble-based approach to improve the accuracy of diabetes diagnosis using machine learning techniques. The proposed model combines the strengths of multiple classifiers to enhance predictive performance across various types of diabetes, including Type 1, Type 2, and non-specific cases. By evaluating key performance metrics such as precision, recall, F-score, and ROC area, the model consistently outperforms individual classifiers and previously published models. The results demonstrate significant improvements in accuracy, with the proposed model achieving up to 96% accuracy for healthy subjects and 95% for both non-specific Type 1 and Type 2 diabetes, highlighting its robustness and reliability in real-world diagnostic applications. The research emphasizes the importance of ensemble learning in addressing the challenges of medical data classification, particularly in terms of reducing false positives and negatives. The proposed model not only enhances diagnostic accuracy but also provides a scalable and efficient solution for early detection of diabetes. This work underscores the potential of advanced machine learning methodologies in healthcare, offering a promising framework for further research into the application of ensemble techniques for diagnosing other diseases and improving patient outcomes.