Comparative Performance Analysis of Machine Learning Techniques for Engagement Assessment in Education

Enhancing education techniques and strategies is crucial to building and developing student knowledge and skills, as it fosters a more engaging learning environment, can accommodate diverse learning styles, and prepares students to meet the demands of a rapidly changing world. Student engagement in classroom discussions and lectures significantly affects the learning outcomes achieved in a course, and with the common ratio of one teacher to a classroom of students, it is challenging for a teacher to determine if all students are actively participating in the learning material. Machine learning can be utilized to adjust and create new learning techniques to capture and hold student attention by detecting and tracking student engagement through social robots. Social Robots have been successfully implemented in many job sectors, including healthcare, therapy, and, most recently, education. To begin integrating social robots into the classroom, a study is needed on the different machine-learning techniques used to train and test the robot’s ability to detect engagement. This study aims to investigate the effectiveness of using a social robot to assess and monitor student engagement during learning activities. For this purpose, this study investigates the performance of various machine learning techniques on publicly available datasets for engagement assessment, which social robots can utilize. The results showed that the dataset is crucial for the machine learning model's improved performance, and CNN models with the selection of the correct activation function performed better.