Enhancing Sequential Recommendation System For MOOCs Based On Heterogeneous Information Networks

This study introduces a novel recommendation framework for MOOCs (Massive Open Online Courses)that leverages HINs (Heterogeneous Information Networks) to enhance sequential recommendation models. This is the first attempt to utilize HINs in this context, offering a more comprehensive approach than traditional methods that rely solely on homogeneous data sources. The framework employs HERec, an attentive meta-path based approach to capture contextual information and learn nuanced representations of learners and items. By combining HERec with sequential recommendation models like BERT4Rec and SASRec, the system provides personalized course recommendations based on user preferences, implicit relationships, and contextual factors, addressing the challenges of data diversity and sparsity. Experiments demonstrate a significant increase in accuracy, with the highest improvement reaching 12.69%, providing learners with more relevant and personalized course suggestions.