An Intelligent Supervised Machine Learning Framework for Diabetes Prediction

Abstract

Estimates suggest that 578 million people could be affected by diabetes mellitus in 2030, with an estimated 700 million affected by 2045 worldwide. Diabetes mellitus (DM) has impacted many people worldwide, with over 422 million affected and about 1.5 million deaths annually. Given the significant global burden of diabetes mellitus and alarming mortality rates, early diagnosis is crucial. Machine learning (ML) techniques offer promising ways for reliably predicting diabetes. This research proposes a machine-learning framework utilizing the random forest algorithm for accurate diabetes prediction. The proposed framework was built using the Hospital Frankfurt, Germany dataset, which included eight independent variables and one target variable. The proposed framework was evaluated and compared to other existing frameworks based on accuracy and AUC metrics for Area Under Curve. The findings reveal that the proposed framework outperformed the reviewed frameworks in this study, with a 99.39% score of accuracy.