A STUDY ON PREDICTION OF DIABETES MELLITUS USING ARTIFICIAL NEURAL NETWORK, CLASSIFICATION AND REGRESSION TREE, LOGISTIC REGRESSION ALGORITHMS

Abstract

Diabetes Mellitus is a chronic metabolic disorder characterized by high blood glucose levels resulting from either insufficient insulin production or the body's inability to effectively use insulin. Early detection and prediction are crucial for effective management and prevention of complications. Artificial intelligence is increasingly being utilized in healthcare for the prediction, management, and diagnosis of various diseases. This study aims to predict diabetes risk using an Artificial Neural Network, Classification and Regression Tree, and Logistic Regression algorithms.
Objective: To predict the risk of diabetes mellitus using Artificial Neural Network (ANN), Classification and Regression Tree (CART), and Logistic Regression algorithms.
Methodology: This case-control study included 400 diabetes patients and 400 healthy controls, recruited from the Medicine OPD of an MGM hospital. A comprehensive dataset, incorporating demographic characteristics, lifestyle factors, and medical history, was collected and used for diabetes prediction. Predictive models—ANN, CART, and Logistic Regression—were developed and validated using cross-validation techniques. Model performance was assessed based on accuracy, sensitivity, specificity, and the area under the receiver operating characteristic curve.
Results: The study found that physical activity, gender, diet, parental history of diabetes, and stress were significantly associated with the prevalence of diabetes. The results showed that the ANN model achieved an accuracy of 82.1%, Logistic Regression achieved 76.0%, and the CART model demonstrated the highest accuracy at 86.8%.
Conclusion: The study highlights that the factors such as physical activity, gender, diet, and stress play a significant role in predicting diabetes risk, with the CART model offering the highest accuracy of 86.8%.