Source Stratified Machine Learning: Balancing Computational Efficiency and Performance in Cardiovascular Risk Prediction

This study analyses models that can work with different computational resources in predicting heart disease. Different model-feature combinations in low, medium and high resource scenarios are evaluated using the dataset obtained from the Kaggle platform and the relationship between performance and resources is analysed. As part of the study, a new metric called “Computational Efficiency Index” (CEI) is proposed that considers both accuracy and computational efficiency. As a result of the analyses, it is found that the combination of robust scaling and KNN provides an F1 value of 0.9082, while the combination of hybrid feature technique and decision tree provides the highest computational efficiency with a value of $\mathbf{1 1 7 0. 5 5}$.