Ensemble Learning Approaches for Cardiovascular Diseases Prediction: A Comparative Evaluation

Cardiovascular diseases (CVDs) continue to be the major cause of mortality worldwide, emphasizing the critical need for early detection methods. Identifying heart disease in its initial stages poses a considerable challenge for healthcare providers. However, recent advancements in diagnostic technologies provide encouraging opportunities for early diagnosis and intervention. This article evaluates the performance of four machine learning algorithms-AdaBoost, XGBoost, LightGBM, and CatBoost-in predicting heart disease across two classes: normal (0) and abnormal (1). The models were assessed based on precision, recall, and F1-score to determine their effectiveness. For class 0, LightGBM achieved the highest F1-score (0.89), with a precision of 0.93 and recall of 0.85, indicating a strong balance between precision and recall. In class 1, AdaBoost, XGBoost, and LightGBM all achieved an F1-score of 0.9, with LightGBM demonstrating higher recall (0.94) and strong overall performance. AdaBoost excelled in recall for class 1 (0.96), while CatBoost had the lowest F1-scores across both classes. The results suggest that LightGBM is the most robust model for heart disease prediction, consistently delivering high precision and recall, making it an effective algorithm for this application.