A Comprehensive Survey of Heart Disease Prediction Approaches: Methods, Applications, Performance Analysis, Datasets, Research Challenges, and Future Scopes

Abstract

Heart disease remains a leading cause of global mortality, accounting for nearly 17.9 million deaths annually. Major risk factors such as hypertension, hyperglycemia, and obesity enable early identification and preventive interventions through lifestyle modifications or medical treatment. Traditional diagnostic methods, including ECGs and coronary angiography, face limitations of inefficiency, invasiveness, or high cost, which points to the importance of non-invasive, reliable, and real-time predictive approaches. Machine learning (ML) and deep learning (DL) have transformed healthcare by enabling decision support systems that analyze clinical parameters for accurate heart disease prediction. This survey provides a comprehensive review of methods for predicting heart disease reported between 2019 and 2025, covering individual classifiers, ensemble techniques, feature selection methods, and state-of-the-art DL architectures. The study evaluates model performance, highlights the role of significant features, and discusses the integration of ML or DL in early detection. The study presents a systematic analysis of publicly available datasets and benchmark studies to guide researchers in model development. Furthermore, the review emphasizes existing research challenges, including target class data imbalance, model generalizability, overfitting, feature redundancy, interpretability, and clinical applicability, while presenting potential solutions and future directions. This study highlights advances in explainable AI to improve transparency and clinician trust. While ensemble and deep learning methods outperform traditional models, challenges such as class imbalance, dataset limitations, and interpretability remain. By consolidating progress across methods, applications, and datasets, this research supports the development of precise, interpreted, and effective frameworks for cardiac disease prediction.