A Blending Ensemble Approach to Predicting Student Dropout in Massive Open Online Courses (MOOCs)

The problem faced in the implementation of Massive Open Online Course (MOOC) is the high dropout rate (DO) reaching 90% which exceeds the formal school dropout rate. Preventive action needs to be taken to minimize the impact on MOOCs, instructors, and students. One solution is to do machine learning...

Full description

Saved in:
Bibliographic Details
Main Authors: Muhammad Ricky Perdana Putra, Ema Utami
Format: Article
Language:Indonesian
Published: Universitas Muhammadiyah Purwokerto 2025-03-01
Series:Jurnal Informatika
Subjects:
Online Access:http://jurnalnasional.ump.ac.id/index.php/JUITA/article/view/24061
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The problem faced in the implementation of Massive Open Online Course (MOOC) is the high dropout rate (DO) reaching 90% which exceeds the formal school dropout rate. Preventive action needs to be taken to minimize the impact on MOOCs, instructors, and students. One solution is to do machine learning (ML) based prediction. The use of ML does not escape the problem of prediction performance that is still less accurate so it needs to be improved by blending ensemble learning (BEL). This research builds a BEL model consisting of two layers including base model with KNN, Decision Tree, and Naïve Bayes algorithms, then meta model with XGBoost. The dataset from KDD Cup 2015 contains clickstream from XuetangX website. The pre-processing stage includes selecting the course with the most participants, normalization, SMOTE, feature selection, and breaking it into three: ensemble, blender, and test data. The BEL model evaluation results obtained an accuracy value of 90.16%, precision of 85.64%, recall of 97.31%, F1-Score of 91.10%, and AUC of 92.83%.
ISSN:2086-9398
2579-8901