Comparative Analysis of Artificial Intelligence Techniques for Cardiovascular Disease Diagnosis and Risk Prediction

Abstract

Cardiovascular disease (CVD) remains the leading cause of global mortality, necessitating accurate and clinically generalizable diagnostic models. Artificial intelligence (AI), particularly machine learning (ML) and deep learning (DL), has demonstrated potential in analyzing clinical data for early CVD detection. However, existing studies predominantly focus on single-model performance within isolated datasets, with limited cross-comparative synthesis. This study provides a comparative evaluation of major AI algorithms for CVD diagnosis and risk prediction. A narrative review of comparative studies published between 2019 and 2026 was conducted using PubMed, Google Scholar, and SciSpace. Performance metrics were synthesized across models including Random Forest (RF), Support Vector Machines (SVM), k-Nearest Neighbors (KNN), Gradient Boosting, ensemble methods, and Deep Neural Networks (DNNs). Results reveal ensemble and Gradient Boosting approaches consistently demonstrated superior predictive performance (AUC: 0.80–0.92). Random Forest frequently outperformed KNN and logistic regression in structured clinical data. Although deep learning models achieved high accuracy in single-cohort studies, limited external validation constrains clinical applicability. This study provides evidence that ensemble-based frameworks currently offer the most effective balance between predictive accuracy, robustness, and clinical feasibility. Future research should emphasize multi-center external validation and explainable AI frameworks.