A Fuzzy Rule-Based System for Classification of Diabetes

Area of Science:

• Medical Informatics
• Computational Intelligence
• Data Science

Background:

• Diabetes is a global health crisis with no cure, making early diagnosis crucial for managing complications.
• Current diagnostic methods often lack interpretability, hindering clinical understanding and trust.
• The rising prevalence of diabetes necessitates advanced, explainable diagnostic tools.

Purpose of the Study:

• To develop an interpretable fuzzy logic-based model for the early diagnosis of diabetes.
• To enhance the explainability of diagnostic processes in diabetes detection.
• To improve the accuracy and reliability of early diabetes identification.

Main Methods:

• Fuzzy logic was integrated with the cosine amplitude method to create two fuzzy classifiers.
• Fuzzy rules were systematically designed based on the developed classifiers.
• A publicly available diabetes dataset was utilized for model evaluation.

Main Results:

• The proposed fuzzy rule-based model achieved a high accuracy of 96.47%.
• The model demonstrated superior performance compared to existing diabetes diagnostic techniques.
• The interpretability of the fuzzy logic approach was successfully maintained.

Conclusions:

• The developed fuzzy logic model offers a highly accurate and interpretable solution for early diabetes diagnosis.
• This approach holds significant potential for application in the healthcare sector for improved patient outcomes.
• The findings support the use of explainable AI in managing chronic diseases like diabetes.