E. Dritsas, M. Trigka, Ph. Mylonas |
EEG-based Engagement Prediction in e-Learning Environments Using Machine Learning Techniques |
Lecture Notes in Business Information Processing, Springer, August 28, 2025 |
ABSTRACT
|
Accurately assessing learner engagement in e-learning environments is crucial for enhancing educational outcomes and optimizing personalized learning experiences. This study presents a machine learning (ML) framework for electroencephalogram (EEG)-based engagement prediction, leveraging multi-channel EEG recordings to capture cognitive responses during learning sessions. A well-defined methodology was implemented, including EEG signals preprocessing, feature extraction based on Power Spectral Density (PSD), and three techniques for feature ranking and selection to identify the most relevant neural features for engagement classification. By evaluating several ML models, including Logistic Regression (LR), Support Vector Machines (SVM), Random Forest (RF), Gradient Boosting Machines (GBM), Neural Networks (NNs), Convolutional NNs (CNNs), and Hybrid Ensemble approach, we demonstrate that feature selection significantly enhances classification performance. The Hybrid Ensemble model achieved the highest accuracy (92.7%) and the area under the ROC curve (AUC) (95.1%) when trained on a highly refined set of 14 features, improving interpretability while reducing computational complexity. The selected features, primarily from temporal, occipital, and parietal EEG channels, align with established neural mechanisms underlying memory processing, sensory integration, and attentional regulation. The results reinforce the potential of EEG-based analytics for real-time engagement monitoring, supporting adaptive e-learning systems that personalize content based on cognitive states.
|
28 August , 2025 |
E. Dritsas, M. Trigka, Ph. Mylonas, "EEG-based Engagement Prediction in e-Learning Environments Using Machine Learning Techniques", Lecture Notes in Business Information Processing, Springer, August 28, 2025 |
[ PDF] [
BibTex] [
Print] [
Back] |