A FUZZY EXPERT SYSTEM FOR EARLY DIAGNOSIS OF DIABETES MELLITUS USING AN ATKINSON INDEX-BASED ALGORITHM

Abstract

Diabetes Mellitus (DM) is a chronic disorder characterized by prolonged high blood sugar due to insufficient insulin production or ineffective insulin use. As a global public health concern, DM affects millions and leads to severe complications such as cardiovascular disease and kidney failure if not properly managed. Early diagnosis is crucial, but is often hindered by overlapping symptoms with other metabolic disorders. Additionally, traditional diagnostic methods, such as fasting glucose and A1C tests, rely on fixed thresholds and fail to account for individual variability, making them less effective. To address these challenges, intelligent systems like Fuzzy Expert Systems (FES) have emerged, effectively managing uncertainty and enhancing early diagnosis. This study proposes a novel FES integrated with the Atkinson Index Algorithm (AIA) for early diabetes diagnosis. The FES utilizes fuzzy logic to handle imprecision and uncertainty in data, while the AIA improves sensitivity in risk assessment by addressing inequality in the distribution of risk factors. The proposed model was evaluated using the PIMA Indians Diabetes dataset and compared with recent studies based on classification accuracy, F1 score, precision, and recall. Results show that the proposed model outperforms the  baseline method as presented in the study, achieving 7.6% higher accuracy, 3.8% higher F1 score, 20.5% higher precision, and 15.5% higher recall. These findings demonstrate the model's effectiveness in diagnosing diabetes while minimizing false positives, establishing it as a more sensitive and reliable diagnostic tool.