Model Pembelajaran Mesin Hibridisasi RFXGB Untuk Prediksi Drop Out Siswa MOOC

Muhammad Ricky Perdana Putra (Universitas Amikom Yogyakarta, Indonesia)

G-Tech : Jurnal Teknologi Terapan
Vol 8 No 4 (2024): G-Tech, Vol. 8 No. 4 Oktober 2024

As a learning supplement or fulfilment of lifelong learners' needs, the non-formal learning platform, MOOC, is quite attractive to the general public. However, there is a problem that 90% per cent of MOOC students experience dropout. As a result, the name of the organiser is affected so that it decreases its reputation. The method that can be done as risk management is to predict dropout students based on Machine Learning (ML) using a hybrid model built with more than one model so as to improve prediction performance. The first layer with Random Forest (RF) algorithm, and the second layer with Extreme Gradient Boosting (XGBoost) algorithm. The dataset is collected from user activity logs on XuetangX MOOC. The results of testing with K-Fold cross validation and confusion matrix, show that hy model gets an accuracy value of 82.89%, precision 85.03%, recall 93.84%, F1-Score 89.22%, and AUC 81.43%.