Attention monitoring of students during online classes using XGBoost classifier

Abstract

With the ever-growing influence of globalization and digitization on education, online learning has gained immense popularity, driven in large part by the global COVID-19 pandemic. In response, educational institutions have rapidly expanded their virtual course offerings and assessment methods. One significant challenge in the realm of online education is the effective assessment of student involvement, engagement, and attentiveness during virtual classes. To address this challenge, we introduce a specialized system tailored to identifying and comprehending the complex range of student behaviors exhibited during online classes. Our system takes into account the diverse contexts and individual considerations that shape students' responses to the online learning environment. It employs user authentication through facial recognition technology and seamlessly integrates vital components, including facial detection, hand tracking, mobile phone detection, and pose estimation modules. Leveraging these components, we extract high-level features to construct a comprehensive dataset for training machine-learning models designed to detect students' attention levels during online classes. After evaluating six prominent machine learning algorithms, we have selected the extreme gradient boosting (XGBoost) algorithm, which demonstrated an impressive 99.75% accuracy on the test data. Finally, the system generates a comprehensive anonymous report on the level of attentiveness of students, accessible through a dedicated webpage that includes a summary of the behavior of students throughout online classes, providing valuable insights for personalized interventions and enhanced online learning experiences.