Abstract

Heart disease causes the greatest number of deaths in world. A large number of people cannot recognize it in early stage. In this study, our goal is to find a good model for prediction of heart disease. Through VIF calculation and Principal Component Analysis, we show that there is no multicollinearity in the data. Then to find the best model, we compare five classification models i.e., Logistic Regression model, Support Vector Machine, Random Forest model, Naïve Bayes classifier and Linear Discriminant Analysis to predict if a person has heart disease or not. We compare the models using 10-fold cross-validation method with three repetitions. The study proposes Random Forest model as the most appropriate predictor of heart disease. The slope of the peak exercise ST segment is the most important subject to predict heart disease. Old peak, type of chest pain and maximum heart rate achieved are also important for predicting heart disease.