DefinePK

Abstract

Risk prediction of heart failure (HF) patients is extremely important to provide them with specific treatment that will assist in enhancing clinical outcomes and quality of life. Globally, cardiovascular diseases are the leading causes of death resulting in almost 17 million deaths annually with most of them being caused by myocardial infarction and cardiac failure. This paper introduces a new Decision-Level Fusion that is powered by Fuzzy Logic (DLFeFL) model to improve predictions of heart failures. The architecture incorporates the outputs of two base learners Support Vector Machine (SVM) and Artificial Neural Network (ANN) into a fuzzy inference system that adds variable adds the decision strengths of the two base learners. In contrast to the traditional ensemble models which use a fixed weighting, the proposed fusion includes an adaptive fuzzy layer that treats the confidence of the classifiers as the linguistic variables that enhances interpretability of decisions and uncertainty. With 299 clinical records, ANN with 91.1% accuracy on 10 hidden layers, and SVM with 86.3% accuracy on 5-fold cross-validation were obtained. DLFeFL fusion performed better than the two base models showing that it is a promising decision-support tool in heart failure prognosis.
Keywords : Heart Failure; AI in Healthcare; Artificial Neural Network (ANN), Support Vector Machine (SVM), Machine Learning.
 
https://doi.org/10.5281/zenodo.17403575