Analysis Kinerja Algoritma Support Vector Machine dan Xgboost Classifier dalam Prediksi Data Penyakit Jantung

Abstract

Heart disease is a common condition in humans and can have negative impacts on health. Various data such as age, gender, blood pressure, and other factors can be used to predict the likelihood of heart disease. This research utilizes medical record data sourced from the Cleveland Clinic Foundation UCI Machine Learning Repository. The author employs the Support Vector Machine and XGBoost Classifier methods for training and testing, along with a 10-fold CV Grid Search to find the best parameters and enhance performance. The classification performance assessment, based on confusion matrix calculations, indicates that the XGBoost Classifier achieves higher accuracy, specifically 98.36% in testing and 85.53% in training, with an execution time of 0.028 seconds. In contrast, the Support Vector Machine attains an accuracy of 93.44% in testing and 83.88% in training, with an execution time of 0.006 seconds. The percentage accuracy difference is 4.92% in favor of the XGBoost Classifier. Furthermore, when employing Grid Search CV 10-Fold, both algorithms obtain optimal parameters and improve performance with an accuracy score of 100%. This demonstrates that both algorithms achieve their best performance optimally. Therefore, based on the significant difference in accuracy when using Grid Search CV 10-Fold and without using it, the proposed XGBoost Classifier algorithm is more accurate and efficient in predicting heart disease data.