Perbandingan Random Forest dan Gradient Boosting pada Prediksi Hasil Belajar Siswa

Abstract

Student learning outcomes are important indicators for evaluating the success of the learning process and can be used to support data-driven academic decision-making. This study aims to compare the performance of the Random Forest and Gradient Boosting algorithms in predicting student learning outcomes. The study employed a quantitative approach with a comparative experimental design using the student dataset containing 500 student records and 11 attributes. The target variable was passed, while several attributes such as gender, age, study_hours_per_week, attendance_rate, parent_education, internet_access, extracurricular, and previous_score were used as predictors. The data preprocessing stage included data cleaning, missing value handling, categorical data transformation, feature selection, and train-test splitting with an 80:20 ratio. Model evaluation was conducted using confusion matrix, accuracy, precision, recall, F1-score, and ROC-AUC, supported by 10-fold cross validation. The results showed that Random Forest slightly outperformed Gradient Boosting on most evaluation metrics. On the test data, Random Forest achieved an accuracy of 87.00%, precision of 90.63%, recall of 89.23%, F1-score of 89.92%, and ROC-AUC of 93.27%, while Gradient Boosting obtained an accuracy of 86.00%, precision of 90.48%, recall of 87.69%, F1-score of 89.06%, and ROC-AUC of 93.10%. These findings indicate that both algorithms performed well in predicting student learning outcomes, with Random Forest showing more stable performance on the dataset used.