Evaluating Machine Learning Models for Predicting Cardiovascular Disease
Author Information
Author(s): Iacobescu Paul, Marina Virginia, Anghel Catalin, Anghele Aurelian-Dumitrache
Primary Institution: Dunărea de Jos University of Galati
Hypothesis
This study aims to evaluate and compare the performance of seven machine learning models for cardiovascular disease prediction.
Conclusion
The k-Nearest Neighbors model achieved the highest accuracy of 99.06%, demonstrating its effectiveness in predicting cardiovascular disease.
Supporting Evidence
- The kNN model achieved an accuracy rate of 99.06%, surpassing previous studies.
- Advanced preprocessing techniques like SMOTE–ENN improved model performance.
- Hyperparameter optimization through Grid Search Cross-Validation enhanced model reliability.
Takeaway
This study tested different computer programs to see which one is best at predicting heart problems, and one program did really well, getting almost all the answers right.
Methodology
The study evaluated seven machine learning models using the BRFSS 2021 dataset, applying advanced preprocessing techniques and assessing performance through metrics like accuracy and AUC.
Potential Biases
Potential bias due to class imbalance in the dataset before applying SMOTE–ENN.
Limitations
The study may have introduced synthetic noise through SMOTE–ENN and relied on predefined parameter grids for optimization.
Participant Demographics
The dataset included clinical and behavioral features relevant to cardiovascular diseases, with a significant class imbalance.
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website