Classification Students’ Attention in E-learning Using Machine Learning

Abstract

Remarkable progress is being made in straddling the boundaries between the actual and the virtual worlds. Computers have become an essential aspect of everyday life as more and more people use them to complete a wide range of jobs, from online learning to shopping. When a user's level of involvement with the system (s)he is dealing with is recognized, the system's response to the user may be altered. As a result, e-learning benefits from cheaper costs and greater value. Machine learning is one of the goals of educational research that focuses on learning behavior analysis. It's a goal of e-learning to help students overcome the downsides of e-learning environments, such as prone inattention during the class, which can degrade student productivity and outcomes. Teachers can get quick feedback and make adaptive adaptations to meet the requirements of their students when they can accurately forecast their engagement. Researchers have made great strides in the field, yet many of their methods rely on the inefficient feature extraction methods or learn global features. Using the Support Vector Machine (SVM) model, we provide a new method for detecting student involvement in this work, called Local-SVM. The local feature extraction is done using the MoveNet model, which selects 17 unique key-points and extracts their values. The proposed model is assessed on the DAiSEE dataset and compared to K-nearest neighbor (KNN) and existing state-of-the-art techniques. This strategy outperforms the competitors by a wide margin, according to evaluation matrices and expert feedback.

Keywords: E-learning, MoveNet, SVM, KNN, DAiSEE

JEL Classification: I2