A Novel Hybrid GSA-MLP Framework for Enhanced Coronary Artery Disease Diagnosis: A Comparative Study with XGBoost and Machine Learning Models

Coronary heart disease remains a critical global health concern, demanding accurate and efficient diagnostic solutions. This research presents a comprehensive evaluation of an advanced Gravitational Search Algorithm-Multilayer Perceptron (GSA-MLP) hybrid model against state-of-the-art machine learning methods for cardiac disease diagnosis. Our enhanced GSA-MLP architecture demonstrates superior diagnostic performance, achieving AY,Y^% accuracy,.,AY F)-Score, and ۹Y AUC-ROC on clinical cardiovascular data. Notably, the proposed GSA-MLP model exhibits exceptional recall of AY,۱%, substantially reducing the risk of false negatives-a crucial factor in medical diagnostics where missed diagnoses can lead to severe consequences. Comparative analysis reveals that GSA-MLP outperforms optimized XGBoost (A.,۰۷% accuracy, VY F-Score) and Random Forest (V۹, %%% accuracy, V۷ F۱-Score) by significant margins. While XGBoost demonstrates superior computational efficiency (*,۲۰ seconds vs. ۱.۷۱,۹ seconds), the diagnostic accuracy and recall superiority of GSA-MLP make it particularly valuable for high- stakes medical applications. The integration of GSA's mass-interaction mechanism enables effective navigation of complex medical decision boundaries, overcoming local optima limitations that constrain conventional neural networks. This research establishes that the evolutionary-neural integration through GSA-MLP represents a substantial advancement in medical diagnostics, demonstrating Yo, %% accuracy improvement over standard MLP and Y,YY% enhancement over XGBoost.