Knowledge-aware graph fusion for community recommendation in edge education

Purpose This study addresses sparse learner profiles and edge-side maintenance constraints in edge education through knowledge-aware community-level resource recommendation. Design/methodology/approach This paper proposes Knowledge-Aware Graph Fusion for Community Recommendation (KAGE-Rec). It uses confidence-gated fusion of educational social relations, learning interests, and knowledge mastery states to construct communities, ranks resources by community interest, prerequisite satisfaction, and knowledge coverage, and updates only affected neighborhoods and communities through local refresh. Findings On MOOCCubeX, KAGE-Rec outperforms all baselines on Recall@10 and normalized discounted cumulative gain@10. For sparse learners, Recall@10 and KCov@10 improve by 9.8 and 13.5%. Knowledge-Aware Graph Community Construction (KAGC) raises weighted modularity from 0.5762 to 0.6034. Local refresh provides up to 13.8 × speedup with a maximum Recall@10 drop of 0.0215. Research limitations/implications Evaluation uses one offline massive open online course dataset without operational edge deployment. Practical implications KAGE-Rec reduces full graph reconstruction under constrained edge resources. Originality/value KAGE-Rec integrates community construction, knowledge-aware recommendation, and edge-side incremental maintenance in one framework.