<![CDATA[Eye diseases such as refractive, implant, cataract, and other disorders can reduce the quality of life of sufferers if not diagnosed and treated properly. Limited access to medical examinations is one of the obstacles in early treatment, so an accurate and efficient technology-based decision support system is needed. This study aims to compare the performance of the Linear Support Vector Machine (SVM) and Modified Balanced Random Forest (MBRF) algorithms applied to handle class imbalance in classifying four categories of eye diseases using medical records of PIM (Prima Inti Medika) Hospital patients consisting of 2,805 data with 30 features. The research process includes preprocessing, model training, and performance evaluation using 5-Fold Cross Validation as well as testing on a test set of 20% of the data. Evaluation metrics used include accuracy, precision, recall, and F1-score. The test results show that MBRF excels in all metrics, with an average F1-Macro cross-validation of 85.67% (3.33% higher than SVM) and a test set accuracy of 87.32% (3.21% higher). Analysis per category shows an average F1 improvement of 3.6%, with the highest improvement in the Refractive class (+5.0%). MBRF also has a 21% faster training time than Linear SVM and provides clinically relevant feature importance information. Based on these results, MBRF is recommended as the main model in a decision support system for diagnosing eye diseases on a similar dataset. ]]>
Copyrights © 2026
