Integrating Deep Learning and Ensemble Techniques for Enhanced Heart Disease Prediction

Cardiovascular diseases (CVD) is amongst the leading causes of the increase in mortality rates worldwide. Various heart-related diseases like heart attacks, strokes, arrhythmias, congenital heart defects, ischemic heart disease, and heart failure has been increasing in the population in recent years. This study highlights the need for early and accurate prediction of heart disease by developing a predictive model formed by the integration of machine learning, deep learning and ensemble techniques in order to ensure early treatment and prevent misdiagnosis. A Kaggle dataset including 70,000 data points and 13 features is used for analysis. The approach of this study is to employ base classifiers like K-Nearest Neighbors (KNN), Random Forest (RF), Support Vector Machine (SVM), Multilayer Perceptron (MLP) and XGBoost (XGB) integrated with deep learning through one-dimensional Convolutional Neural Network (1D-CNN). These models are further combined using ensemble techniques like stacking over K-Nearest Neighbors (KNN), Random Forest (RF) Support Vector Machine (SVM) and a voting classifier to give the final prediction with both hard and soft voting methods. The model’s performance is evaluated based on standards like accuracy and area under the ROC curve (AUC). SVM performed the best amongst all the base classifiers by attaining a maximum accuracy of 88.02% and AUC 0.95, followed by MLP with an accuracy of 87.85%, XGBoost with 87.83%, RF with 87.77% and lastly KNN with an accuracy of 86.64%. Furthermore, the stacking model gave an accuracy of 87.92%. The voting classifier employed on the model to achieve the best accuracy of 87.98% under hard voting and the highest AUC of 0.96 under soft voting.