A Comprehensive Performance Analysis of Various Classifier Models for Coronary Artery Disease Prediction

Abstract

Cardio vascular diseases (CVD) are the major reason for the death of the majority of the people in the world. Earlier diagnosis of disease will reduce the mortality rate. Machine learning (ML) algorithms are giving promising results in the disease diagnosis, and they are now widely accepted by medical experts as their clinical decision support system. In this work, the most popular ML models are investigated and compared with one other for heart disease prediction based on various metrics. The base classifiers such as support vector machine (SVM), logistic regression, naïve Bayes, decision tree, k-nearest neighbour are used for predicting heart disease. In this paper, bagging and boosting techniques are applied over these individual classifiers to improve the performance of the system. With the Cleveland and Statlog datasets, naive Bayes as the individual classifier gives the maximum accuracy of 85.13%and 84.81%, respectively. Bagging technique improves the accuracy of the decision tree, which is identified as a weak classifier by 7%, and it is a significant improvement in identifying CVD. that Bayes, Support Vector Machine and Logistic are strong classifiers more than 80% accuracy and Decision Tree and K Nearest Neighbours as weak classifiers. Bagging and boosting techniques the performance of weak classifiers Decision Tree and K Nearest Neighbours. Bagging technique improved the accuracy of the decision tree algorithm 7.77% maximum for Statlog dataset. In future, feature selection is to be applied to find out the most relevant features of the data set and applying over the ensemble models over it will give better-improved accuracy.