Comparative Evaluation of Ensemble Machine Learning Approaches for Heart Disease Prediction.

Area of Science:

• Computational biology
• Machine learning
• Data science

Background:

• Heart disease prediction is crucial for public health.
• Ensemble Learning offers robust predictive modeling.
• Benchmarking algorithms aids in selecting optimal methods.

Purpose of the Study:

• To computationally benchmark Ensemble Learning algorithms for heart disease prediction.
• To compare hard voting, soft voting, and stacking ensemble methods.
• To evaluate the impact of data preprocessing on model performance.

Main Methods:

• Utilized a publicly available cardiovascular dataset (1,190 instances, 11 features).
• Applied data preprocessing: handling missing values, outlier removal, scaling, class balancing.
• Feature selection using Random Forest (RF); ensemble models: Decision Tree, Random Forest, AdaBoost, XGBoost; meta-model: Logistic Regression.

Main Results:

• Stacking ensemble classifier achieved the highest accuracy of 91.88% on the test dataset.
• Comparative analysis included precision, recall, and F1-score.
• Accuracy was the primary criterion for comparing individual and combined classification systems.

Conclusions:

• The stacking ensemble classifier demonstrates superior performance in heart disease prediction.
• The study provides a systematic protocol for comparing data preprocessing and ensemble configurations.
• Emphasis is on methodological evaluation rather than clinical validation.