Advancing Diabetes Risk Prediction Using Anfis: a Fuzzy-Logic Approach for Improved Healthcare Decision Support

Fuzzy logic and machine learning integration has shown promise in healthcare analytics, particularly for chronic disease prediction. This study leverages the Adaptive-Network-based Fuzzy Inference System (ANFIS) to assess and predict the risk of Type 2 diabetes. The model is trained and evaluated using the “Pima Indians Diabetes Database,” a widely used dataset in medical research, enabling effective analysis of key health indicators associated with diabetes onset. Gaussian membership functions manage data uncertainty, improving predictive accuracy. The model achieved 85.42 % training accuracy and 81.30 % test accuracy, with F1-scores of 78.44 % and 74.72 %, respectively. Performance trends across epochs were analyzed, and learned membership functions visualized. The results highlight ANFIS's potential for early diagnosis and personalized treatment in healthcare.