Abstract
Due to the complex nature of cardiovascular disease and the complexity of early symptoms, early and effective detection of cardiovascular disease continues to be a major issue in clinical practice. In order to increase diagnostic accuracy and resilience, this study suggests a hybrid ensemble learning architecture that combines Machine Learning (ML) and Deep Learning (DL) models. The method uses two DL models, Feed forward Neural Networks (FNN) and Simple Recurrent Neural Networks (RNN), in addition to four ML classifiers, a K-Nearest Neighbour (KNN), a random forest (RF), Decision Tree (DT) and Extreme Gradient Boosting (XGB). In order to ensure cleaner and more dependable input data, the system also includes sophisticated pre-processing, such as outlier detection utilizing Isolation Forest and Modified Z-Score techniques. The benefits of base learners are combined using a weighted voting ensemble technique based on stacking. The suggested ML-DL ensemble outperforms individual classifiers and traditional model ensembles with an accuracy of 94.22%, according to experimental evaluation using a publicly accessible Kaggle heart disease dataset. The findings verify that integrating ML and DL into a single ensemble structure greatly improves model stability, prediction reliability, and applicability for early cardiovascular disease identification.
Keywords: Cardiovascular Disease, Deep Learning, Ensemble Learning, Outlier Detection, Stacking Model, Weighted Voting.