Stacked Ensemble Learning for Early Heart Disease Prediction Using PCA–LDA and SMOTE

Heart disease continues to place a heavy burden on global health, making early and dependable diagnosis extremely important. This study introduces an ensemble based learning approach aimed at improving the reliability of heart disease prediction. The proposed framework integrates two feature reduction techniques, Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA), to simplify the data while retaining the most informative patterns. Because real clinical datasets often contain far fewer disease positive cases, the Synthetic Minority Oversampling Technique (SMOTE) is applied during preprocessing to create a more balanced and representative dataset. The model incorporates a diverse set of classifiers, including Random Forest (RF), Support Vector Machine (SVM), k Nearest Neighbours (kNN), Logistic Regression (LR), Decision Tree (DT), and Gaussian Na¨ıve Bayes (GNB). Building on these models, the main contribution of this work is a Stacking Based Ensemble Learning Model (SELM) that employs a meta learner to integrate the strengths of these classifiers while reducing their weaknesses. The proposed model achieves a maximum accuracy of 96.8%, along with high F1 scores and AUC values, indicating balanced and stable predictive performance. These findings highlight the strong potential of the SELM framework as a reliable component in computer-aided cardiovascular risk assessment systems.