Prediction of Diabetes Using Data Mining and Machine Learning Algorithms: A Cross-Sectional Study

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Diabetes Research

Background:

• Early diagnosis and treatment of diabetes are crucial for improving patient outcomes and quality of life.
• Predictive modeling offers a promising approach for identifying individuals at risk of diabetes.
• Machine learning and data mining techniques can enhance the accuracy of diabetes risk prediction.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting fasting blood glucose status.
• To identify the most significant risk factors associated with diabetes using data mining techniques.
• To support the planning of diabetes management and prevention strategies through accurate risk assessment.

Main Methods:

• Cross-sectional study of 3376 adults from a diabetes screening program in Tehran, Iran.
• Dataset balancing using random sampling and synthetic minority over-sampling technique (SMOTE).
• Feature selection via Shapley values and noise analysis; evaluation of five ML algorithms (CatBoost, random forest, XGBoost, logistic regression, ANN) using accuracy, sensitivity, specificity, F1-score, and AUC.

Main Results:

• Age, waist-to-hip ratio, body mass index, and systolic blood pressure identified as key predictors of fasting blood glucose status.
• All models demonstrated similar predictive abilities, with the CatBoost model achieving the highest Area Under the Curve (AUC) of 0.737.
• Gradient boosted decision tree models effectively identified significant diabetes risk factors.

Conclusions:

• Machine learning, particularly gradient boosted decision trees like CatBoost, can accurately predict diabetes risk.
• Key modifiable and non-modifiable risk factors include age, waist-to-hip ratio, body mass index, and systolic blood pressure.
• The developed model can serve as a valuable tool for public health initiatives in diabetes management and prevention.