A hybrid metaheuristic optimization and machine learning framework for smart healthcare based cardiovascular disease prediction

Abstract

Cardiovascular diseases (CVDs), particularly heart attacks, remain among the most serious global public health challenges, causing millions of deaths annually and imposing substantial pressure on healthcare systems worldwide. Early identification of at-risk individuals is therefore essential for timely intervention and improved clinical outcomes. In recent years, machine learning (ML) has emerged as a promising tool for supporting medical decision-making through the analysis of complex clinical and lifestyle data. However, the predictive effectiveness of ML models is often limited by the presence of redundant, irrelevant, or noisy features, which can reduce model accuracy, increase computational complexity, and hinder interpretability. To address these challenges, this study proposed an optimization-driven feature-selection framework to enhance the predictive performance of ML-based heart disease diagnosis systems. The proposed approach systematically integrated metaheuristic optimization algorithms with classical ML classifiers to identify the most informative feature subsets from cardiovascular datasets. Two widely used benchmark datasets, namely the CVD dataset and the Heart Attack dataset, were employed to evaluate the effectiveness of the framework. Six state-of-the-art metaheuristic optimization algorithms were investigated for feature selection, including the Bat Algorithm (BA), Whale Optimization Algorithm (WOA), Jaya Algorithm (JA), Particle Swarm Optimization (PSO), Firefly Algorithm (FA), and Flower Pollination Algorithm (FPA). The optimized feature subsets were subsequently evaluated using three well-established ML classifiers: K-Nearest Neighbors (KNN), Naïve Bayes (NB), and Decision Tree (DT). The experimental results demonstrated that integrating metaheuristic feature selection significantly improved classification performance compared with baseline models trained using the full feature space. Among the evaluated optimization methods, the Jaya Algorithm consistently produced the most effective feature subsets across both datasets and achieved the highest predictive performance. In particular, the JA-based models achieved classification accuracies of up to 98.0% on the CVD dataset and 93.4% on the Heart Attack dataset, outperforming other optimization strategies as well as conventional ML models. To demonstrate practical applicability, the developed predictive model was integrated into a prototype mobile application that enabled preliminary heart risk assessment based on user-provided health information. By combining optimized feature selection, ML-based prediction, and a mobile-enabled healthcare interface, the proposed system provided a scalable and accessible solution for early cardiovascular risk screening. Overall, the study demonstrated that metaheuristic-driven feature optimization can significantly enhance the reliability and accuracy of ML-based cardiovascular disease prediction systems, thereby contributing to the advancement of intelligent diagnostic solutions within the emerging Healthcare 4.0 ecosystem.