Comparative Analysis of AI/ML Models for Student Performance Prediction Using XuetangX MOOC Dataset

Author Information

Mahmoud Yousef AlFaress, Lincoln University College, Malaysia
Midhunchakkaravarthy Janarthanan, Lincoln University College, Malaysia
Chandra Kumar Dixit, Institute of Engineering and Technology, India
Omar Al Jadaan, Ras Al Khaimah Medical and Health Sciences University, United Arab Emirates
Satheesh Babu, Lincoln University College, Malaysia
Shashi Kant Gupta, Chitkara University Institute of Engineering and Technology, India

Abstract

In this paper, we provide a thorough study on the AI/ML models for predicting student performance in MOOCs using the XuetangX MOOC dataset. An early detection of students-at-risk is important due to the increasing use of online learning systems in order to produce better learning results. We compare six models – Random Forest, Support Vector Machine, Logistic Regression, Convolutional Neural Network, Long Short-Term Memory, and Transformer –in their predictive ability for student dropout in MOOCs. This comparative analysis is unique due to its focus on the large-scale XuetangX MOOC dataset, providing insights into model performance across diverse feature sets and offering practical implications for early warning systems in a specific regional context. Our findings, however, demonstrate that traditional machine-learning models outperform deep learning approaches (AUC 0.58-0.63 vs. 0.50-0.56), with Logistic Regression achieving the highest performance while maintaining better interpretability. Feature importance analysis reveals that course progress rate, quiz success rate, and session duration play the most significant roles in student success prediction. We further demonstrate the practical application of these models in an early warning system that provides as-recorded personalized interventions. The results illuminate the trade-offs between model complexity, performance, and interpretability in Educational Data Mining (EDM), and have important implications for learning analytics researchers and educational technologists interested in developing AI-supported student support systems.

Save as a Favourite