Prediction of Cardiovascular disease using machine learning algorithms on healthcare data

Abstract

Cardiovascular Disease (CVD) is a leading cause of death worldwide, with the potential to cause serious conditions such as heart attacks and strokes. Early assessment of CVD can significantly reduce mortality rates. In recent studies, machine learning algorithms have been applied to Electronic Health Records (EHR) to estimate risk factors for myocardial infarction. This article explores the use of various machine learning techniques on a healthcare dataset to predict a 10-year risk of future coronary heart disease (CHD). The dataset used in this study was obtained from the Framingham and Massachusetts cardiovascular study. We found that our models achieved varying levels of accuracy: 64% for logistic regression, 83% for Naïve Bayes classifier, 42% for Support Vector Machine (SVM), 65% for Random Forest, 78% for KNN classifier, and 70% for XGBOOST classifier. It is revealed that a patient with no history of heart disease may benefit from an algorithm such as  Naive Bayes Classifier, while an older patient with a history of heart disease may require an algorithm such as Support Vector Machine. These factors can help guide the physician in selecting the most appropriate algorithm for each individual patient, ensuring that the diagnosis is as accurate as possible and that the treatment plan is tailored to meet the patient's unique needs.