Prediction of heart diseases using majority voting ensemble method

Abstract

Heart disease is the one of the most serious problems in healthcare and affects large number of people. It is very important to detect it on time, otherwise it can cause serious consequences, such as death. In this paper, we applied artificial neural network, k-nearest neighbor and support vector machine algorithms to build model which will be used for prediction of heart disease. Multiple experiments using each of these algorithms are performed. Additionally, the ensemble learning is applied, and results are compared. Initially, the problem was approached as multiclass classification, however it was transformed into the binary classification problem, to simplify model since number of outputs is reduced from five to two. In both cases, the highest accuracies are obtained by majority voting which are 61.16% for multiclass classification and 87.37% for binary classification.