Analisis prediksi mortalitas pada pasien gagal jantung menggunakan perbandingan algoritma support vector machine dan k-nearest neighbors

Abstract

Heart failure is one of the leading causes of mortality worldwide, requiring accurate prediction methods to support early detection and medical decision-making. This study compares the performance of Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) algorithms in predicting mortality among heart failure patients. The dataset used is the Heart Failure Clinical Records Dataset, consisting of 299 patient records with 13 clinical features, including age, ejection fraction, creatinine level, hypertension, anemia, and smoking status. The research process involved data preprocessing, feature standardization, model training, and evaluation using accuracy, precision, recall, F1-score, and balanced accuracy. The results indicate that SVM achieved superior performance with 0.80 accuracy, 0.84 precision, 0.70 recall, and 0.72 F1-score, while KNN reached only 0.68 accuracy and 0.52 F1-score. These findings demonstrate that SVM is more reliable in detecting mortality risk in heart failure patients, especially under imbalanced data conditions. This research provides valuable insights for the development of machine learning-based medical decision support systems, enabling healthcare professionals to perform earlier and more targeted interventions.