Multi-model Ensemble Based Approach for Heart Disease Diagnosis

Heart is the core part responsible for blood pumping in the human body. Any disorder that effectuates the functions of the cardiovascular named heart disease or cardiovascular disease. CVD (cardiovascular disease) turns into a major cause of death worldwide. This cardio disease leads to 17.9 million deaths per year according to an estimation by World Health organization. Machine learning approaches become the effective for prediction of heart disease to overcome the mortality rate. Researchers deployed various techniques to predict heart disease. Proposed ensemble methods that provide the optimal model by combining multiple models. Stacking, bagging, voting, and boosting are the ensemble methods that have been deployed on the UCI dataset of heart disease by feature extraction. Feature scaling is performed using a standard scalar to normalize the feature set or independent variables. Data is segmented into training and testing set in the ratio of 70:30 for machine learning models. Experiments results indicated that bagging-XGB (extreme gradient boosting) achieved 88% which is the highest accuracy among all proposed models. In boosting methods AdaBoost achieved a high accuracy of 86% among all proposed techniques. The soft voting classifier acquired an accuracy of 86% which is more than the hard voting classifier. A combination of Random Forest, Gaussian Naive Bayes, and K-Nearest Classifier achieved the highest accuracy of 87% among all others stacking approaches. All ensemble techniques are evaluated by the ROC curve (receiver operating characteristic curve) and performance measures. Experiments results indicate that heart disease can be effectively predicted by ensemble techniques.