Heart Disease Prediction Using Multi-Model Ensemble Learning

Cardiovascular diseases (CVDs) represent a major global public health challenge. Early identification of high-risk populations and timely interventions are critical strategies for reducing disease burden. With advancements in big data technologies, machine learning (ML) and artificial intelligence (AI) have emerged as pivotal tools for cardiac risk prediction. However, single-model approaches face limitations in handling complex medical data. This study proposes a multi-model ensemble learning framework tailored for clinical practice, leveraging the 2020 Behavioral Risk Factor Surveillance System (BRFSS) dataset from the U.S. Centers for Disease Control and Prevention (CDC). The dataset comprises <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{3 1 9, 7 9 5}$</tex> samples and <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{1 8}$</tex> features. Through optimized feature engineering, diversified base model selection (e.g., Random Forest, XGBoost, Logistic Regression), and integration of stacking fusion with Bayesian hyperparameter optimization, we developed an efficient hybrid ensemble framework. Experimental results demonstrate superior performance of the stacking fusion model on the test set, achieving an ROC-AUC of 0.9297, accuracy of 85.29 %, and F1-score of 86.57 %. The model significantly reduces misclassification rates for both positive and negative cases, offering a reliable tool for clinical decision support.