Predictive Modelling of Glycated Hemoglobin Levels Using Machine Learning Regressors

Abstract

Diabetes is a chronic condition characterized by elevated levels of blood glucose, also known as hyperglycemia. Measurement of HbA1c is a widely used blood test that provides an essential tool for monitoring diabetic progression and assessing the effectiveness of diabetes management but this test is usually not conducted until there are some symptoms of diabetes in the patient and sometimes it goes unnoticed for a longer period resulting in the late detection of the disease. This study proposes a novel approach to HbA1c Prediction using machine learning regression algorithms on various features including Age, BMI, and hematological parameters. This study also compares the performance of ten machine learning regressors on the prediction of HbA1c level using performance metrics such as Mean square error, Root mean squared error, Mean absolute error MAE, R square, Adjusted R square, and Minimum Absolute Percentage Error. Result: Linear regression was found as the best performer with an R square and adjusted R square value of 1.00, Mean square error, Root mean squared error, Mean absolute error, and Minimum Absolute Percentage Error of 0.00. A higher HbA1c Level predicted using this method should go for actual HbA1c testing for confirmation.