An accurate fuzzy rule-based classification systems for heart disease diagnosis

Physicians and healthcare providers need to better understand the thought processes and methods used in clinical decision-making. This allows physicians to diagnose and detect diseases early, especially heart disease that causes death. The diversity and availability of healthcare data encourage clinicians to use healthcare applications in the diagnosis process. Most of these applications use machine learning techniques to make accurate and fast decisions. On the other hand, Explainability in healthcare applications increase the level of clinician confidence and reduces the risk of making wrong decisions, thus expands the scope and efficiency of healthcare applications. In this paper, we propose a novel data-driven method based on fuzzy clustering and linguistic modifiers to design a fuzzy rule-based classification system for heart disease diagnosis. The proposed system provides an interpretable knowledge base to explain the decision-making process. Regarding the experiment, we have used Cleveland, Hungarian and Va long beach heart disease datasets to compare the proposed method with five known machine learning methods for predicting heart disease: Artificial neural network, Support Vector Machine, K-Nearest Neighbor, Naïve Bayes, and Random Forest. The findings show that the proposed model is superior in terms of balancing interpretability and precision.