Comparing Machine Learning Models for Heart Disease Prediction

One of the top causes of death globally is heart disease. Each year, an estimated 17.9 million people die due to heart disease, contributing to 31 percent of all deaths worldwide. Heart diseases, particularly cardiac arrest, could happen anytime and anywhere, without prior warnings or indications. Thus, being able to predict if heart disease is present in a patient can help both the patients and doctors be aware of a potential cardiac arrest and take necessary precautions. Early prognosis of heart disease can essentially help in effective and preventive treatments of patients and reduce the risk of complication of heart disease. In this study, a machine learning approach is used on clinical data of patients to learn models for the prediction of heart disease in patients. A correlation study of the features in the data was carried out to support feature selection for the study. Then, a comparative study of five machine learning techniques, namely Logistic Regression, Naïve Bayes, K-Nearest Neighbour, Decision Tree and Support Vector Machine, was conducted to compare the performance of the models for heart disease prediction. The results obtained were from 13 clinical parameters used to learn models for predicting heart disease. Logistic Regression seemed to perform comparatively well compared the other techniques