Performance-optimized boosted hybrid ensemble model for diabetes risk prediction

Abstract

The proposed boosted hybrid ensemble (BHE) machine learning (ML) model utilizes the classification power which reduces the overfitting by bagging and generates better results using random forest (RF) and extreme gradient boosting (XGBoost). The paper presents the importance and impact of secondary features in type 2 diabetes prediction utilizing real-time self reported and hospital data. The research study shows that age, gender, body mass index (BMI), and glucose are the key prime factors and are also influence by the other factors like demographic conditions, eating, and activity styles to some extents. The paper presents transfer learning (TL) on the basis on standard Pima Indians diabetes dataset (PIMA) to apply hybrid 2-layer BHE model to predict and classify the records into diabetic and non diabetic class providing explanations to factors contributing to it. The result section shows the highest 98% accuracy for BHE with optimized model presenting recommendations as per careful considerations of World Health Organization (WHO) and American Diabetes Association (ADA) standards. The paper throws light on the need of life-style factors considerations and correction to establish causation and refine preventive strategies in avoiding or postponing type-2 occurrences in youth people. This paper present perfect integration of multifactorial data with high reliability of artificial intelligence (AI)-driven healthcare explainable models to generate recommendations utilizing TLs.