Tuberculosis (TB) is one of the infectious diseases that remains a global health problem, requiring an accurate computational approach to assist in the prediction and clinical decision-making processes. This study aims to evaluate the performance of Support Vector Machine (SVM) in predicting tuberculosis and analyze the contribution of Information Gain (IG)-based feature selection to improving model performance. The dataset used was obtained from the Mendeley public repository and consisted of 493 patient data with 16 encoded attributes. The research stages included data preprocessing, calculation of Information Gain values, selection of relevant features, training of the SVM model with Radial Basis Function (RBF) kernel, and performance evaluation using accuracy, precision, recall, and F1-score metrics. The results of the experiment show that the SVM model without feature selection achieved an accuracy of 0.99, with a recall value of 0.97 for the negative class and 1.00 for the positive class. The application of Information Gain produced identical performance, namely an accuracy of 0.99, a macro-average F1-score of 0.99, and a weighted-average F1-score of 0.99. Confusion matrix analysis, both in absolute and normalized forms, showed a very low classification error rate in both scenarios. These findings indicate that the dataset has highly informative features, so feature selection based on Information Gain is not necessary.
Copyrights © 2026