<![CDATA[Class imbalance in datasets poses a significant challenge in the application of classification models, including the Classification and Regression Tree (CART) method. This study aims to evaluate the performance of CART combined with two data balancing techniques: Random Under Sampling (RUS) and Synthetic Minority Oversampling Technique (SMOTE). The data set used in this research is the Heart Failure Clinical Records from Kaggle.com, which exhibits an imbalance where the number of deceased patients is 1,568 records (minority class) and the number of survivors is 3,432 records (majority class), with a total of 5,000 records. The RUS technique reduced the total number of records to 2,526, with each class containing 1,263 records. Conversely, after applying SMOTE, the total number of records increased to 5,474, with each class containing 2,737 records. Model performance evaluation was conducted using precision, recall, and F1-score metrics, both before and after implementing data balancing techniques. The results of the study showed that combining CART with SMOTE produced better performance in recognizing the minority class compared to RUS, achieving accuracy and F1-score of 88.203% and 88.195%, respectively. Meanwhile, RUS achieved an accuracy of 86.345% and an F1-score of 86.332%. Therefore, the use of SMOTE improved model accuracy by approximately 1.85% and F1-score by 1.86% compared to RUS. This study makes a significant contribution to improving prediction accuracy on imbalanced datasets and enriches scientific references related to the application of the CART method and data balancing techniques.]]>
