Empowering adaptive MOOCs: A learning path recommendation model based on multi-level knowledge graphs and rule-guided path constraint optimization

Massive Open Online Courses (MOOCs) provide abundant learning resources, yet their fixed instructional structures and limited adaptive support constrain learning effectiveness. Previous path recommendation approaches typically depend on predefined start and target nodes, offer limited dynamic adaptability, and insufficiently consider multi-hop dependencies among learning content. To address these challenges, this study proposes an adaptive learning path recommendation model based on a multi-level knowledge graph and rule-guided path constraint optimization (KGRRec). First, we design a multi-level knowledge graph framework that models course knowledge structures and integrates MOOC resources across three levels: instructional organization, knowledge representation, and resource aggregation. Second, building on this framework, we develop a rule-guided path constraint optimization method, which incorporates dynamic reasoning mechanisms, path constraint modeling, and learning gain simulation to deliver adaptive learning path recommendations. Implemented in a prototype integrating seven database-related MOOCs, KGRRec was evaluated through a one-group pretest-posttest study (34 students) and a randomized controlled study (59 students). Results showed KGRRec significantly outperformed eight baselines on Prerequisite, CogValidity-F, and expert ratings. Furthermore, it showed a trend toward improved learning performance and supported positive learning experiences. These findings provide preliminary evidence of the model’s effectiveness and a methodological reference for the future development of large-scale adaptive MOOCs.