MACHINE LEARNING APPROACHES FOR EARLY DIABETES PREDICTION: A COMPARATIVE STUDY USING CLINICAL DATA

Abstract

This study focuses on the accurate and early prediction of diabetes, which plays a vital role in improving clinical care and treatment planning. The research evaluates the performance of three machine learning techniques - Logistic Regression, Random Forest, and Support Vector Machine (SVM) - using data collected from 769 patients in Chennai. The models were compared based on their ability to predict diabetes effectively. The results indicate that the SVM model achieved the highest prediction accuracy of 78%, while Logistic Regression and Random Forest recorded accuracies of 77% and 76% respectively. Statistical analysis showed that the variation in performance among the models was not significant (p = 0.3484). Even though SVM demonstrated slightly better predictive capability, Logistic Regression provided clearer interpretation of risk-related factors, making it more suitable for clinical decision-making. In addition, the Random Forest model helped identify the most influential variables associated with diabetes risk. Overall, the findings suggest that machine learning approaches can support early diagnosis and assist healthcare professionals in developing personalised treatment and management strategies for diabetic patients. Incorporating a wider range of clinical attributes in future studies may further improve prediction performance and enhance medical risk assessment.