Related Experiment Videos
Optimizing predictive performance in heart disease diagnosis with stacked wrapper pre-processing techniques
1Department of Computer Science and Engineering, Koneru Lakshmaiah Education Foundation, Aziz Nagar, Hyderabad, Telangana 500075, India.
Computational Biology and Chemistry
|August 7, 2026
Summary
This study introduces a Stacked Wrapper Attribute Machine Learning Model (SWA-ML) to improve heart disease prediction accuracy by addressing missing data and class imbalance in clinical datasets. The SWA-ML model enhances feature estimation and classification for better risk assessment.
Area of Science:
- Cardiovascular Health
- Machine Learning in Medicine
- Data Science
Background:
- Heart disease prediction relies on machine learning models using patient data, but real-world datasets often suffer from missing information and class imbalance, compromising accuracy.
- Biased models may inaccurately predict the majority class (no heart disease), necessitating advanced techniques for reliable risk assessment.
Purpose of the Study:
- To develop and evaluate a novel Stacked Wrapper Attribute Machine Learning Model (SWA-ML) for enhanced heart disease prediction and classification.
- To address challenges of missing data and class imbalance in heterogeneous clinical datasets for improved predictive performance.
Main Methods:
- Integration of three public heart disease datasets (1800 instances) with preprocessing including attribute alignment, missing-value imputation, encoding, scaling, and class balancing.
- Implementation of wrapper-based feature selection (RFE, Forward Selection, Backward Elimination, Genetic Algorithm) and stacked ensemble architecture within the SWA-ML framework.
- Rigorous model evaluation using stratified train-test partitioning, k-fold cross-validation, hyperparameter optimization, and comparison with existing methods.
Main Results:
- The SWA-ML model demonstrated consistently higher classification accuracy, precision, recall, F1-score, and AUC compared to individual feature selection methods.
- Ablation experiments, statistical significance testing, and calibration evaluation confirmed the efficacy and robustness of the proposed framework.
- The stacked wrapper configuration proved superior in handling complex clinical data for heart disease forecasting.
Conclusions:
- The SWA-ML model offers an effective and interpretable decision-support tool for predicting heart disease risk using structured clinical data.
- This approach successfully mitigates issues of missing data and class imbalance, leading to more reliable predictions.
- The study highlights the potential of ensemble methods and wrapper-based feature selection in advancing cardiovascular risk prediction.