Abstract
The medical diagnosis often dealt with uncertainty and vagueness that hindered the effectiveness of conventional ML approaches. This limitation was overcome by the integration of the LDFS with ML algorithms in this study on heart disease diagnosis. The LDF framework is a powerful structure that has reference parameters that can easily change the physical meaning of its attributes. Our proposed hybrid model eliminates the need for pipelines like in conventional ML for handling categorical and numerical features, as it accommodates both feature types through membership functions. Several ML algorithms, like logistic regression, decision tree, support vector machine, and XGBoost, were evaluated on both the crisp dataset and LDF-based datasets. A comparative analysis demonstrates that our proposed LDF-ML consistently outperforms conventional ML algorithms in classifications. All the performance metrics were increased on LDF-Datasets by 0.97% in accuracy, 0.95% in precision, 0.99% in recall, and 0.97% in F1 score for the XGBoost Algorithm. Thus, the proposed integration provides a new direction for medical diagnosis as well as decision-making in terms of handling ambiguity with improved interpretability.