Enhancing the Understandability and Interpretability of Type-II Diabetes Diagnosis Using a Deep Fuzzy Neural and XAI Approach

Deep fuzzy neural applications have been shown to be effective in improving the accuracy of type-II diabetes diagnosis. However, due to their inherent complexity, deep fuzzy neural applications are often not easy to explain and communicate between doctors and patients. In addition, in existing explainable artificial intelligence (XAI) applications, the explanations generated by different XAI techniques have not been aggregated, leading to user confusion or contradictory interpretations. Aiming to tackle this issue, this study devises a deep fuzzy neural and XAI approach that leverages various XAI techniques simultaneously and aggregates the explanations to resolve conflicts. A real case from NIDDKD was analyzed to validate the deep fuzzy neural and XAI approach. According to the experimental results, the deep fuzzy neural and XAI approach elevated the diagnosis accuracy by more than 18%.