Enhancing Cardiovascular Disease Diagnosis: The Power of Optimized Ensemble Learning

Cardiovascular disease, a leading cause of global mortality, is influenced by genetic, environmental, and lifestyle factors. Accurate diagnosis involves identifying key risk factors such as age, gender, family history, diabetes, obesity, and sedentary lifestyle, using medical history, physical exams, and lab tests. This study introduces an advanced method called GMM-based Dynamic Ensemble Learning Optimization using a Multi-Objective Evolutionary Algorithm based on Decomposition (MOEA/D). This approach utilizes evolutionary resampling inspired by the Gaussian mixture model (GMM) to enhance classifier diversity, performance, and quantity, producing optimal datasets. Ensemble weight interpolation and dynamic ensemble techniques are applied to reduce overfitting. Tested on seven heart disease datasets with varying imbalance rates, the proposed method consistently surpasses established ensemble algorithms across multiple metrics, validated by McNemar’s and Wilcoxon tests. Additionally, Kaplan Meier Estimates analysis of a heart failure dataset provides deeper insights into heart disease prognosis. This research advances accurate heart disease prediction and prognosis through innovative data mining and ensemble learning techniques.