Development and Validation of a Novel Bayesian Belief Network: A Reliable Fuzzy Weighted Diabetes Predictive Model

The rising burden of chronic diseases, particularly diabetes, necessitates diagnostic frameworks that can navigate the inherent ambiguity of clinical data. Conventional predictive models often struggle with the stochastic uncertainty stemming from subjective patient narratives and laboratory noise, as well as the 'black-box' lack of interpretability. To transcend these limitations, this research introduces a novel Fuzzy Weighted Bayes Association Rule Mining (FWBARM) framework. This approach integrates fuzzy logic to handle data vagueness with a weighted mechanism that implicitly learns feature importance, thereby generating robust, transparent, and clinically interpretable decision rules. The proposed system, evaluated as the 'Reliable Diabetes Prediction Model,' demonstrated superior diagnostic efficacy, achieving 96.8% accuracy, 98.6% precision, and 97.5% recall. By reconciling high predictive performance with rule transparency, this work offers a scalable solution for personalised medicine and reliable Clinical Decision Support Systems.