ELECTROENCEPHALOGRAM SIGNALS FOR DETECTING CONFUSED STUDENTS

Abstract

Online education has become a vital learning medium, especially during the COVID-19 pandemic. However, its lack of face-to-face interaction poses challenges in assessing students' engagement and understanding. This study addresses this issue by utilizing electroencephalogram (EEG) data to detect student confusion on massive open online course (MOOC) platforms. A novel feature engineering technique, Probability-Based Features (PBF), is introduced to enhance machine learning model performance. We employ three machine learning models—Support Vector Classifier (SVC), K-Nearest Neighbors (KNN), and Recurrent Neural Network (RNN)—to classify EEG data into confused and non-confused categories. By leveraging probability-based feature engineering, the system enhances classification performance and enables a more accurate detection of student confusion. The models are trained and evaluated using EEG data collected from students interacting with online learning modules.