Computational Comparative Review-Driven Analysis on Machine Learning Classifiers for Heart Disease Detection

The accurate prediction of heart disease is essential for timely diagnosis and effective clinical intervention. Advances in machine learning have facilitated the development of predictive models that employ both conventional classifiers and advanced hybrid or deep learning frameworks. This study selects a total of 39 research papers published between 1 January 2008 and 31 December 2025 for the computational comparative review of machine learning models for myocardial infarction and heart disease through their classifiers, feature selection methods, data preprocessing techniques, and optimization strategies. Support Vector Machines, Random Forests, and Artificial Neural Networks have emerged as the most effective individual classifiers, whereas hybrid and ensemble approaches consistently deliver superior accuracy, sensitivity, and robustness. Feature selection techniques, including Particle Swarm Optimization, Genetic Algorithms, Recursive Feature Elimination, and Relief, in conjunction with data balancing methods such as Synthetic Minority Over-sampling Technique, are critical for enhancing predictive performance, albeit with increased computational demands. Deep learning architectures, notably Convolutional Neural Networks (CNN) and Multiple Feature Branch CNN, demonstrate exceptional predictive capability, frequently achieving accuracies above 98%, particularly on real-world clinical datasets. Integrating optimized feature selection, balanced data handling, and advanced classifiers represents the most effective strategy for constructing reliable, high-performance heart disease prediction systems, underscoring future directions in precision healthcare and clinical decision support.