Abstract: This study presents a fuzzy logic–based diagnostic system for the early detection of Diabetes Mellitus, aimed at improving diagnostic accuracy and interpretability in the presence of uncertain clinical data. Traditional diagnostic techniques, such as threshold-based glucose and HbA1c evaluations, often fail to capture the gradual transition between normal and diabetic conditions. To address this limitation, a Mamdani-type Fuzzy Inference System (FIS) was developed using input parameters including fasting blood glucose, HbA1c, BMI, age, and family history. The model converts crisp clinical data into linguistic variables (Low, Normal, High) and applies a structured rule base to evaluate diabetes risk levels. Implementation was carried out using MATLAB’s Fuzzy Logic Toolbox and Python’s scikit-fuzzy library, with validation performed using the Pima Indians Diabetes Dataset. The system achieved high diagnostic performance with an accuracy of 90.5%, sensitivity of 92.4%, and specificity of 88.7%, demonstrating its efficiency and reliability. The results indicate that fuzzy logic provides a robust and human-like reasoning framework for medical decision-making, making it an effective tool for early diabetes diagnosis and clinical decision support.

Keywords: Fuzzy Logic, Diabetes Mellitus, Medical Diagnosis, Fuzzy Inference System, Early Detection, Mamdani Model, Decision Support System

Downloads: | DOI: 10.17148/IARJSET.2025.124123