Penerapan Algoritma Random Forest untuk Klasifikasi Tingkat Keparahan Penyakit pada Data Rekam Medis

Abstract

Accurate determination of disease severity is an important step in supporting medical decision-making. This study aims to classify the severity of patients’ diseases into three categories—Mild, Moderate, and Severe—using the Random Forest algorithm. The data used were obtained from patients’ medical records containing structured clinical parameters and have undergone a preprocessing stage, including data cleaning, variable transformation, and splitting into training data (80%) and testing data (20%). The test results show that the Random Forest model achieved an accuracy of 74.77%. The best performance was obtained in the Mild class with a recall value of 0.95 and an f1-score of 0.84. The Moderate class achieved a recall of 0.71 and an f1-score of 0.73, while the Severe class showed perfect precision (1.00) but a low recall (0.12), indicating the model’s limited ability to detect cases in this class. The macro average values for precision, recall, and f1-score were 0.83, 0.60, and 0.59 respectively, while the weighted average values were 0.78, 0.75, and 0.71 respectively. These findings indicate that Random Forest can be used to classify disease severity based on medical records with relatively good performance for the majority class, but further optimization—such as data balancing or parameter adjustment—is needed to improve sensitivity toward classes with fewer samples.