SGFRec: Personalized Course Recommendation Based on Sequence Graph Fusion

Accurately modeling learners’ course preferences is a key task in Massive Open Online Course (MOOC) recommendation. Existing sequential recommendation methods can capture recent changes in learners’ interests but have limited ability to exploit shared preference patterns across learners. In contrast, graph-based recommendation methods can capture collaborative information from learner–course interactions, but they often ignore the effects of behavior order and time intervals on interest evolution. To address these problems, this study proposes Sequence Graph Fusion Recommendation (SGFRec), a method for MOOC course recommendation. SGFRec models learners’ recent interests from course-taking sequences and collaborative preferences from learner–course interactions. It further considers the time intervals between course interactions to distinguish different learning rhythms. An adaptive fusion mechanism then balances recent individual interests and collaborative information for each learner. Experimental results on the MOOCCube and MOOCCourse datasets show that SGFRec outperforms several baselines, demonstrating improved offline recommendation ranking performance.