Detecting Learning Patterns and Student Engagement in Online Courses Using Deep Learning

This study introduces LearnTrans, a novel model
architecture that integrates the Transformer architecture
with attention mechanisms to discern learning patterns and
gauge student engagement in online courses. LearnTrans
employs Transformer encoder layers with self-attention
mechanisms to capture dependencies within the sequential
interactions of students with course content. Through
rigorous experimentation on a diverse dataset collected from
prominent online learning platforms, including Coursera,
Udemy, and edX, LearnTrans demonstrates significant
performance improvements over baseline methods.
Specifically, the model achieves an average accuracy
increase of 33% in learning pattern detection and 29% in
student engagement prediction tasks. These findings
underscore the efficacy of the proposed LearnTrans model
in capturing intricate patterns and dependencies within
online learning data, offering promising avenues for
enhancing educational outcomes in digital learning
environments.