Performance Analysis of Base and Meta Classifiers and the Prediction of Cardiovascular Disease using Ensemble Stacking

Abstract

According to World Health Organization, heart disease is the principal cause of death. In the medical domain, to improve diagnosis accuracy researchers have introduced several data mining techniques for the prediction of cardiovascular diseases. The aim of the proposed research is that prediction of heart disease more precisely using an ensemble stacking model which is based on the mixing of heterogeneous classifiers. The research article consists of major two parts. First, analysis on choosing of best meta classifier with a different set of base classifiers and secondly, prediction using an ensemble framework. The experimental end prediction compared with other data mining algorithms. Further, the performance analysis is carried out by accuracy, precision, and recall and f1 score. Better analysis was done by ROC, P_R curve, and AUC. Analysis of the ensemble result shows that Ensemble techniques give better accuracy of 90.16% for testing dataset. Precision, Recall and f1 scores for 92%, 85% and 88% for the classification of sick patients, whereas 89%, 94% and 91 % for healthy patients. The AUC is 0.88 for the heart disease dataset.