Machine learning algorithms for heart disease diagnosis: A systematic review

Abstract

Background

The heart is a vital organ that pumps blood throughout the body. Its proper functioning is crucial for maintaining overall health, and any malfunction can significantly impact other bodily systems. Recently, machine learning has emerged as a valuable tool in cardiology, enhancing the prediction and diagnosis of heart diseases. By analyzing clinical data, these algorithms reveal patterns that traditional methods might miss, aiding in early detection and personalized treatment. This study aimed to evaluate the most widely used and accurate supervised machine-learning algorithms for predicting and diagnosing heart disease.

Methods

A systematic analysis was conducted using research articles obtained from six reputable academic databases: Scopus, PubMed, ScienceDirect, Dimensions, ProQuest, and IEEE. The review covers the years from 2013 to 2024. The focus was on the application of various supervised machine-learning algorithms for diagnosing heart disease.

Result

The study identified twenty-four relevant studies that examined the use of supervised machine learning algorithms for diagnosing and predicting heart disease. Among these, five algorithms were prominent: Decision Trees, Logistic Regression, Naive Bayes, Random Forests, and Artificial Neural Networks. Decision Trees were found to be the most commonly applied and best-performing algorithm, followed by Logistic Regression and Naive Bayes. However, Artificial Neural Networks and Random Forests received less attention despite their potential for high accuracy in certain contexts.

Conclusion

The research findings highlight important trends in heart disease prediction models using supervised machine learning. By examining these trends, researchers can identify algorithms that improve forecasting accuracy, guiding future research objectives and advancing the effectiveness of heart disease diagnosis.