Heart Sync Ensemble for Heart Disease Prediction Using Ensemble Learning with SMOTE

Heart disease continues to be a primary global health concern, requiring accurate and interpretable predictive systems for timely diagnosis. This research introduces HeartSync Ensemble, a novel framework designed for heart disease prediction using advanced machine learning techniques. The framework integrates ensemble learning methods, including stacking and voting classifiers, and a hybrid Convolutional Neural Network (CNN) with Random Forest (RF) to enhance predictive accuracy. To address class imbalance, the Synthetic Minority Oversampling Technique (SMOTE) is applied. Further, K-best feature selection is employed to identify the most influential predictors and improve model performance. K-fold cross-validation ensures robust model evaluation. The proposed approach is validated using publicly available heart disease datasets. Additionally, Explainable Artificial Intelligence (XAI) techniques are incorporated to interpret model decisions, making the system transparent and suitable for clinical decision support. Comparative results, both pre- and post-feature selection, demonstrate improved accuracy, reliability, and interpretability of HeartSync Ensemble.

Keywords: Heart Disease Prediction, Ensemble Learning, SMOTE, CNN-Random Forest, Feature Selection, XAI, K-fold Cross-Validation, Clinical Decision Support.