Improving the accuracy of diagnosing and predicting coronary heart disease using ensemble method and feature selection techniques

Heart disease is a complex disease, and many people around the world suffer from this disease. Due to the lack of a healthy lifestyle, it is the most common cause of death worldwide. Machine learning plays an important role in medical treatment. The goal of this research is to develop a machine learning model to help diagnose heart disease quickly and accurately. In this article, an effective and improved machine learning method is proposed to diagnose heart disease. We designed a novel and robust ensemble model that combines the top three classifiers, namely Random Forest, XGBoost and Gradient Boosting Machine, to effectively diagnose heart disease. We used an ensemble voting method to combine the results of the top three classifiers to improve the prediction of heart disease. We used a combined heart disease dataset containing five different datasets (Hungary, Statlog, Switzerland, VA Long Beach and Cleveland). Feature selection algorithms (Pearson Correlation, Univariate Feature Selection, Recursive Feature Elimination, Boruta Feature Selection, Random forest, and LightGBM) are used to select highly relevant features based on rankings to improve classification accuracy. The proposed ensemble model is designed using seven highly relevant features, and a comparison of machine learning algorithms and ensemble learning techniques is applied to the selected features. Different performance evaluation methods are used to evaluate the proposed model: accuracy, sensitivity, precision, F1-score, MCC, NPV and AUC. Results analysis shows that the ensemble model achieves excellent classification accuracy, sensitivity, and precision of 96.17%, 98.37%, and 94.53%. Our proposed model performs better than existing models and individual classifiers. The results show that the proposed ensemble method can effectively predict the risk of heart disease.