Abstract:
Online learning is growing in popularity these days. As a result, students typically contribute millions of course- related responses to discussion forums and exchange learning experiences. This study focuses on evaluating student’s concentra- tion level in online courses using deep learning architecture and identifying how this affects students’ ability to stay connected and achive effective outcome offered by different MOOC platforms. This experiment proposes a unique method to address the issues by evaluating students’ concentration levels using Deep Learning architecture. This method involves capturing videos of students through their webcams and analyzing the video frame by frame. The goal of this study is to determine whether the issue lies with students’ concentration, the course material, or both. The objectives of our research include evaluating video data, measuring concentration levels, comparing model performances, providing class-based concentration levels (attentive, inattentive, and sleepy), and identifying key factors. We created the dataset ourselves. It underwent preprocessing, which included resizing for analysis, frame extraction, and annotation for classification. Through experimentation, the proposed models can classify different concentration states with up to 92% accuracy. Our research provides educators with insightful information to help increase the overall effectiveness of online learning. Moreover, the study advances the field by offering a systematic approach to assessing and evaluating students’ concentration in online courses.

Index Terms — Deep Learning, Concentration Level, Online Learning, MOOC, ConcentrateNet.