Framework for Early-Stage Diabetes Mellitus Risk Prediction Using Hybrid Supervised Learning

Diabetes mellitus, shortly termed as diabetes is one of the prevalent chronic diseases in the world, which results in high mortality rate. Diabetes is a deadly disorder, which in turn increases the risk of other diseases such as stroke, diabetic neuropathy, diabetic retinopathy, kidney disease, and heart diseases. Therefore, early detection as well as treatment of diabetes can save the human lives to a greater extent. Further, in the existing literature, only popular machine learning techniques such as support vector machine (SVM) are explored for the diabetes risk prediction problem. However, hybrid supervised learning techniques with robust nature such as support vector regression are not much explored in the literature. Furthermore, the existing diabetes prediction methods are less focusing on performance optimization metrics such as R-squared error, which in turn significantly enhance the prediction performance of the the given diabetes detection system. To handle these issues, this article introduces a Diabetes Mellitus Risk Prediction (DMRP) framework for predicting early-stage diabetes by employing hybrid machine learning techniques followed by the accuracy optimization. The experiments on real-world patient datasets and detection results in terms of F1-score, accuracy, and recall metrics clearly prove the efficiency of the given DMRP framework.