Computational Health Informatics for Diabetes Risk Prediction: An Optimized Fuzzy Logic Framework

Abstract

In order to provide a non-categorical substitute for conventional screening methods, this study sought to create an optimum fuzzy logic system to evaluate type 2 diabetes risk for the Iraqi population. The Simulated Annealing heuristic was used to enhance the membership functions of a Mamdani-type fuzzy inference system. The system's performance was assessed using diagnostic metrics in comparison to the existing FINDRISC questionnaire after it was trained and tested on clinical data from 295 patients in Baghdad. Results showed that the optimised fuzzy system matched the gold standard's sensitivity (87.5%) but achieved superior specificity (69.41% vs. 52.55%) and improved positive predictive value (0.3097 vs. 0.2244). The conclusion affirms that the simulated annealing-optimised fuzzy system is a viable and more descriptive tool for personalised diabetes risk prediction, effectively distinguishing between healthy and at-risk individuals based on clinical profiles.