The untapped promise of educational data on social media: A machine learning approach to unveiling online learners

This is the authors' abstract. We don't add key points for this paper.

Massive Open Online Courses (MOOCs) attract millions of learners who engage in discussions about online learning on social media, creating a rich source of educational big data. However, current educational data mining often overlooks the broader learner demographics of learners across multiple courses. This study bridges this gap by analysing Weibo posts from 2013 to 2022 related to the MOOC phenomenon. Using machine learning and deep learning algorithms, learners were identified, with TextCNN proving the most effective. The Latent Dirichlet Allocation (LDA) topic model categorised post content, and Dynamic Topic Models (DTM) tracked the evolution of these topics over time. This research provides a novel perspective on online learners through social media, offering insights into diverse learning communities, identifying often-overlooked silent learners, and presenting a comprehensive view of the online learning experience. The findings lay a crucial foundation for integrating social media data into educational big data to improve online teaching and learning outcomes.