The two-tier machine learning framework combining CNN feature extraction with CatBoost and LightGBM classifiers optimized via metaheuristics achieved binary cardiovascular disease risk prediction accuracy slightly over 92% on a large publicly available cross-sectional dataset.
Does a hybrid machine learning framework combining CNNs and gradient boosting classifiers accurately predict cardiovascular disease risk?
A hybrid machine learning framework combining CNNs and gradient boosting classifiers achieved >92% accuracy in predicting cardiovascular disease risk.
Effect estimate: Accuracy 92% achieved with best models using two-tier framework with optimized hyperparameters
Accurately assessing a patient’s likelihood of developing cardiovascular conditions is essential for proper case classification and for ensuring timely, targeted medical intervention. To address this need, the present study employs a carefully optimized machine learning framework to predict such risks within cardiology settings. A hybrid architecture is proposed that combines convolutional neural networks (CNNs) with cutting-edge gradient boosting classifiers, namely CatBoost and LightGBM, whose performance is further enhanced by metaheuristic optimization. The system adopts a two-layer design capable of capturing complex data structures while supporting accurate classification of cardiac patients and their risk of developing cardiovascular disease. Extensive evaluation on real-world data confirms the framework’s effectiveness for binary classification, with the best models reaching an accuracy of slightly over 92%. To complement predictive performance, explainable AI methods were applied to clarify model decisions, yielding practical insights that can guide future data collection strategies and improve diagnostic precision.
Villoth et al. (Thu,) conducted a other in Adults from a large-scale cardiovascular risk dataset including demographic, lifestyle, anthropometric, and health indicators with class imbalance representing cardiovascular disease risk status. Two-tier machine learning framework combining CNN for feature extraction and CatBoost or LightGBM gradient boosting classifiers optimized by metaheuristic algorithms (including adapted Variable Neighborhood Search) vs. Baseline machine learning models without metaheuristic optimization or classical VNS optimization was evaluated on Predictive accuracy for binary classification of cardiovascular disease risk (Accuracy 92% achieved with best models using two-tier framework with optimized hyperparameters). The two-tier machine learning framework combining CNN feature extraction with CatBoost and LightGBM classifiers optimized via metaheuristics achieved binary cardiovascular disease risk prediction accuracy slightly over 92% on a large publicly available cross-sectional dataset.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: