Heart Disease Detection Scheme Using a New Ensemble Classifier

Abstract

AbstractVarious machine learning techniques are being implemented to deal with heart diseases because of their better performance and their ability to understand the relationship between various input and output features when compared to an experienced physician/doctor. These input features usually involve the values of various tests done on an individual. Many classification, clustering, and deep learning algorithms are being implemented by researchers from all around the world, but after looking at the increase in the rate of heart diseases every year, some more unexplored methods should be implemented along with the application of few techniques so as to enhance the results on the existing classification algorithms. In this paper, a new method of majority voting ensemble learning has been developed to identify the existence or non-existence of heart disease in an individual. In ensemble learning, various classification algorithms are combined to form a more powerful and an accurate algorithm. The algorithms such as logistic regression (LR), naïve Bayes (NB), decision tree (DT), random forest (RF), support vector machines (SVM), Xtreme gradient boosting machine (XGBM), light gradient boosting machine (LGBM), and K-nearest neighbors (KNN) have been used. Each algorithm was individually implemented on the dataset, and based on the evaluated parameters for all of the algorithms, two ensembles were formed by combining the best algorithms. The constructed ensembles are Ensemble 1 consisting of SVC and LR and Ensemble 2 consisting of NBC, RF, SVC, and LGBM with an accuracy of 88.33% in both the ensembles. Other parameters like sensitivity, specificity, precision, and negative predictive value (NPV) were also calculated for every model.KeywordsClassificationHeart diseaseEnsemble learningROC curveMachine learning