Synergistic fusion of optimized ensemble algorithms and deep neural paradigms for precision stratification and prognostication in cardiovascular pathologies

Abstract

Heart disease remains one of the leading causes of mortality globally, necessitating the development of accurate predictive models to assist in early diagnosis and treatment planning. This paper presents a novel machine learning framework for cardiovascular diagnostics using the UCI Heart Disease dataset. The framework integrates advanced feature engineering techniques, including polynomial feature interactions and Random Forest-based feature selection, to enhance model performance. We employed various machine learning models such as Logistic Regression, Support Vector Machines (SVM), XGBoost, and Random Forest, alongside a Convolutional Neural Network (CNN), optimized through the hyperparameter opti-mization tool, Optuna. The Random Forest model achieved the highest accuracy of 97%, demonstrating superior performance over the other models. CNN, though traditionally used for image data, was effectively applied to structured medical data, achieving an accuracy of 96.20%. These results emphasize the robustness of ensemble learning and deep learning models in clinical prediction tasks. Our approach shows how integrating feature engineering and model optimization can improve the accuracy and reliability of heart disease prediction, providing valuable insights for clinical decision support systems.