AN OPPOSITIONAL CROW SEARCH OPTIMIZER FOR OPTIMIZING ENSEMBLE APPROACH WITH PCA-BASED DIMENSIONALITY REDUCTION FOR HEART DISEASE PREDICTION

Heart disease remains a leading cause of mortality worldwide, requiring accurate and efficient predictive systems for timely diagnosis. In this study, we propose a novel Adaptive Oppositional Crow Search Optimizer (AOCSO) integrated with Principal Component Analysis (PCA) and ensemble learning for heart disease prediction. The method employs opposition-based learning to accelerate convergence and dynamically balance exploration and exploitation during hyperparameter optimization. PCA reduces redundancy, while the ensemble framework enhances robustness and generalization. Experimental evaluations were performed on two benchmark datasets: the Cleveland Heart Disease dataset (UCI) and the Framingham Heart Study dataset (Kaggle). Results demonstrate that AOCSO consistently outperforms state-of-the-art baselines—HEXAI, RST-ML, SXG, OEPCA, and EHO—across Accuracy, F1-score, MCC, Specificity, and AUC. On the Cleveland dataset, AOCSO achieved an Accuracy of 96.1% (80:20 split) and 95.5% (70:30 split), surpassing the strongest baseline (OEPCA) by up to 1.1% in Accuracy and 0.02 in MCC. Similarly, on the Framingham dataset, AOCSO obtained 86.2% (80:20 split) and 85.6% (70:30 split), improving MCC by up to 0.02–0.03 over baselines. These results validate the effectiveness of AOCSO in optimizing ensemble models for clinical decision support.