Optimizing Diabetes Care Through Fuzzy Rule-Based Models and Machine Learning Algorithms

Reliable medical decision-making is crucial for early disease diagnosis, particularly for diabetes, which can lead to severe health complications if untreated. The availability of diverse healthcare datasets enables the integration of computer applications in medical diagnostics. However, many datasets lack comprehensive information, impacting diagnostic accuracy. Fuzzy logic effectively addresses vagueness and uncertainty in medical data and serves as a robust model for developing diagnostic systems. This study uses a simulated fuzzy diabetes dataset validated by medical professionals to evaluate ten supervised machine learning approaches—Multilayer Perceptron (MLP), Random Tree, J48, Naïve Bayes Updatable, Bagging, Bayes Net, Logistic, AdaBoostM1, Stacking, and Random Forest. Performance measures such as precision, F1-score, accuracy, recall, and confusion matrix were analyzed. For eight algorithms the accuracy showed 100%, while AdaBoostM1 and Stacking achieved 79.82% and 67.89% respectively. The fuzzy model demonstrates high accuracy, making it an effective decision-support tool to detect diabetes and taking care in the healthcare sector.