Advanced Diabetes Prediction Using Supervised Machine Learning Technique: Random Forest

Abstract

Diabetes remains one of the major causes of untimely death globally. Over 11% of the global population is diabetic, possibly due to late disease detection, inadequate interventions, and lifestyle choices etc. The growing severity of diabetes is driving scientific interest in leveraging Digital Health Technologies (DHTs) for improved management and treatment. Early diagnosis of diabetes is essential for effective interventions, reducing complications, and lowering the mortality rate associated with the disease. Thus, this study focuses on prediction of diabetes using supervised machine learning technique, specifically Random Forest Algorithm (RFA) for timely detection and prevention of the disease. The model was trained using Pima Indian dataset (diabetes), which is freely available on Kaggle database. Trial result indicate that the model was promising, with an accuracy of 92%, 89% precision, 88% recall, and a 90% F1-score. The study shows that applying the Random Forest algorithm significantly improves the accuracy and efficiency of early diabetes detection and diagnosis. However, in spite of the prospects of ML models in diabetes management, there are still concerns about its drawbacks including algorithmic bias, legal and ethical issues, and implementation challenges in clinical environment. Thus, we recommend that legal framework should be put in place to guide the use ML algorithms, and other digital health technologies in clinical diabetes care delivery.