Abstract:
Diabetes mellitus is a chronic metabolic disorder that affects millions of people worldwide and has
been related to serious complications such as cardiovascular disease, neuropathy, nephropathy,
and retinopathy. Effective diabetes care is dependent on maintaining glycemic control, which is
frequently measured using glycated hemoglobin (HbA1c) values. HbA1c is an important
biomarker that represents average blood glucose levels over the previous two to three months and
is used to determine long-term glycemic management and the risk of problems.
The goal of this research is to create a reliable and accurate HbA1c prediction model utilizing
machine learning (ML) technologies. The objectives include assessing clinical and demographic
data to find significant variables, creating and comparing multiple machine learning models, and
determining the best effective model for HbA1c prediction. In addition, the study investigates the
clinical consequences of precise HbA1c forecasts, as well as the obstacles and ethical problems
involved with machine learning in health care.
The study indicated that Gradient Boosting Regression (GBR) outperformed other models, with
the lowest Mean Squared Error (MSE) and greatest R² values. Random Forest Regression also
fared well, although Neural Network Regression (NNR) and Support Vector Regression (SVR)
were less effective due to their sensitivity to feature scaling and parameter adjustment. Accurate
HbA1c forecasts can assist healthcare practitioners anticipate levels and enhance glycemic control,
but issues like individual variability and data security must be addressed.
