<![CDATA[Heart disease remains one of the leading causes of death worldwide, making early detection efforts essential to minimize more serious risks. This study compares four machine learning algorithms—Random Forest, CatBoost, LightGBM, and XGBoost—to determine which model is most effective in predicting heart disease risk. The dataset used was sourced from Kaggle, comprising a total of 918 data points and 12 clinical features related to cardiovascular conditions. The research process included data pre-processing, class balancing using SMOTE, data partitioning, model training with hyperparameter tuning, and evaluation using various performance metrics. The results showed that Random Forest had the highest discriminatory ability with an AUC value of 0.9385. CatBoost, on the other hand, showed the most stable performance with an accuracy of 0.91 after tuning, and had balanced precision and recall in both classes. LightGBM and XGBoost also provided competitive results, although they were still slightly below the two best models. Overall, this study shows that ensemble methods such as Random Forest and CatBoost have great potential for use as decision support in detecting heart disease earlier and more accurately]]>
Copyrights © 2026
