Supervised Machine Learning-Based Models for Predicting Raised Blood Sugar.

Area of Science:

• Medical Informatics
• Public Health
• Machine Learning

Background:

• Raised blood sugar (hyperglycemia) is a significant indicator of prediabetes and diabetes mellitus, a prevalent non-communicable disease (NCD) with increasing global rates.
• Early detection of diabetes is crucial for preventing severe health complications, yet large-scale screening programs face cost and resource challenges.
• The rising prevalence of diabetes, particularly in developing nations, necessitates innovative and efficient detection methods.

Purpose of the Study:

• To develop and evaluate supervised machine learning models for the early detection and prediction of raised blood sugar.
• To identify key diabetes risk factors for model development, including age, BMI, lifestyle, and existing conditions.
• To compare the performance of various classification algorithms in predicting hyperglycemia.

Main Methods:

• Utilized a dataset from the STEPwise approach to NCD risk factor study in the Palestinian community, focusing on adults.
• Employed supervised machine learning algorithms: Random Forest, Decision Tree, Adaboost, XGBoost, Bagging Decision Trees, and Multi-Layer Perceptron (MLP).
• Input features included diabetes-related risk factors such as age, BMI, eating habits, physical activity, other diseases, and fasting blood sugar.

Main Results:

• The Random Forest classifier achieved the highest accuracy at 98.4% in predicting raised blood sugar.
• Bagging Decision Trees, XGBoost, MLP, AdaBoost, and Decision Tree models also demonstrated high performance, with accuracies ranging from 94.8% to 97.4%.
• The models effectively utilized diabetes risk factors for accurate prediction, highlighting the potential for data-driven screening.

Conclusions:

• Supervised machine learning models, particularly Random Forest, offer a highly accurate and efficient approach for detecting and predicting raised blood sugar.
• These models can support early identification of individuals at risk for prediabetes and diabetes mellitus, facilitating timely interventions.
• The findings suggest a promising role for machine learning in enhancing diabetes screening strategies, especially in resource-constrained settings.