Comparative Analysis of Machine Learning Techniques for Predicting Diabetes

Abstract

Diabetes, a chronic illness causing serious health problems, affects millions of people globally. With cases expected to rise, effective strategies for managing, detecting, and preventing the disease are essential. Artificial intelligence (AI) and machine learning (ML) have become powerful allies in this fight. These advancements aid in the automated detection of eye complications (retinopathy), supporting clinical decisions, identifying high-risk populations, and empowering patients to manage their health. The significant public health challenge of diabetes in Zimbabwe, impacting all demographics, highlights the need for better solutions. This research aims to develop a precise predictive model for diabetes using the CRISP-DM methodology. Machine learning techniques like random forest, Naive Bayes, XGBoost, decision trees, and support vector machines, were used to predict the presence of diabetes. The results revealed that the random forest approach outperformed other models, demonstrating a larger area under the curve (AUC).