A Novel Link Prediction Approach for MOOC Forum Thread Recommendation Using Personalized PageRank and Machine Learning

MOOCs are online learning platforms that offer free and open educational resources for global learners. Its online forums enable learners to interact with instructors and peers and thus improve their learning outcomes and social presence. However, learners may struggle to find posts that match their goals and preferences due to the large scale and diversity of MOOCs. A well-designed MOOC forum thread recommendation system can help learners overcome information overload and improve their learning experience. In this paper, we consider the recommendation of forum threads as a link prediction problem and propose a novel link prediction approach based on various Heterogeneous graph feature extraction methods with LightGBM. We conduct experiments on real-world data from a MOOC platform to evaluate the effectiveness of our model. Results show that our approach outperforms various baselines in terms of precision and recall. Our findings demonstrate the effectiveness of combining PageRank with machine learning and our model’s potential to increase student engagement in learning discussion and foster socio-collaborative learning.