An improved ensemble learning approach for the prediction of heart disease risk
Abstract
Heart disease is the leading cause of death globally, and early detection is crucial in preventing the progression of the disease. In this paper, an improved machine learning method is proposed for the prediction of heart disease risk. The technique involves randomly partitioning the dataset into smaller subsets using a mean based splitting approach. The various partitions are then modeled using classification and regression tree (CART). A homogeneous ensemble is then created from the different CART models using an accuracy based weighted aging classifier ensemble, which is a modification of the weighted aging classifier ensemble (WAE). The approach ensures optimal performance is achieved. The experimental results on the Cleveland and Framingham datasets achieved classification accuracies of 93% and 91%, respectively, which outperformed other machine learning algorithms and similar scholarly works. The receiver operating characteristic curves further validates the improved performance of the proposed ensemble learning approach. The results show that heart disease risk can be predicted effectively by the proposed ensemble.