A Hybrid Fuzzy Logic-Machine Learning Framework and Application to Healthcare Risk Prediction

Abstract

Artificial intelligence (AI) and machine learning (ML) have achieved remarkable performance across diverse domains, but many models still struggle with uncertainty, vagueness, and interpretability. Fuzzy logic, originating from Zadeh’s seminal work on fuzzy sets, provides a mathematical framework to model linguistic concepts and gradual membership, making it a natural complement to data-driven ML methods. In this paper, we propose a hybrid fuzzy logic-machine learning framework that integrates fuzzy inference with supervised learning to handle uncertainty in both features and decision boundaries, while improving interpretability. We formalize the proposed method using fuzzy sets, fuzzy rules, and a learnable defuzzification layer, and we present an end-to-end training algorithm. As a case study, we apply the framework to healthcare risk prediction, where input variables such as blood pressure and symptom severity are naturally described by linguistic terms. The experimental results on a benchmark healthcare dataset demonstrate that the proposed fuzzy-enhanced neural network achieves competitive accuracy with improved interpretability compared with conventional ML models. We discuss the advantages, limitations, and potential extensions of the framework, and we provide a comparative analysis with existing hybrid fuzzy-ML approaches. The paper aims to bridge soft computing concepts with modern ML practice and to motivate further work on dynamic and data-driven fuzzy systems.