<![CDATA[The achievement of learning outcomes in programming courses remains a challenge in higher education due to variations in students’ logical thinking skills, problem-solving abilities, and practical competencies. Conventional evaluation methods are generally retrospective and do not provide early identification of students at risk of not achieving course learning outcomes. Therefore, predictive modeling based on educational data can support data-driven academic decision-making. This study aims to develop a predictive model of learning outcomes in a programming course using the Random Forest algorithm combined with feature selection to improve model performance and interpretability. This research employed a computational experimental method with a quantitative approach. The dataset consisted of 180 student academic records, including assignment scores, quizzes, practicum, project, attendance, midterm exam, and final exam scores. The experiment compared a baseline Random Forest model using all features with a model applying feature selection based on feature importance. Data were divided into 80% training and 20% testing sets and evaluated using accuracy, precision, recall, and F1-score. The results showed that the baseline model achieved 83.33% accuracy, while the model with feature selection improved accuracy to 88.89% and increased recall performance. Final exam and practicum scores were identified as the most influential predictors. The findings indicate that integrating Random Forest and feature selection enhances prediction accuracy and provides meaningful insights for early intervention strategies in programming education.]]>
Copyrights © 2025
