Ensemble Based Machine Learning Model for Heart Disease Prediction

The World Health Organization reports that more than 10 million people worldwide die each year due to heart disease. In this project, an ensemble-based machine learning (ML) model is proposed for the prediction of heart disease. The proposed method was evaluated on two datasets: UCI heart disease and Framingham heart study. Various ML such as K-Nearest Neighbors, Naïve Bayes, Decision Tree, Support Vector Machine (SVM), Logistic Regression (LR) and Random Forest (RF) were carried out to predict the outcome of heart disease. Additionally, the Stacking and Voting ensemble approaches are also employed for heart disease prediction. All ML classifiers evaluated on both datasets have their performance metrics analyzed and compared. The results of all ML classifiers assessed on the Framingham dataset is inferior than that of the UCI heart disease dataset. For individual ML classifier, SVM achieved the highest accuracy for UCI heart disease with 91.09%, whereas LR achieved the highest accuracy for the Framingham heart study with 85.38%. However, it should be noted that RF trained on the UCI dataset is the best ML for prioritizing Sensitivity (92.52%). Finally, with the exception of Sensitivity, the Voting approach is strongly suggested because it outperforms the other classifiers noticeably where it achieved Accuracy (91.96%), F1 (91.69%), Sensitivity (91.72%), Specificity (90.77%), and Precision (92.40%) on UCI dataset.