Abstract
The rapid expansion of online education has intensified the need for precise and efficient course recommendation systems to address challenges like information overload and personalised learning. This study bridges three critical research gaps: (1) the lack of integration between knowledge management and deep learning in educational recommendations, (2) the persistent trade-off between recommendation accuracy and computational efficiency and (3) insufficient modelling of pedagogical relationships in existing systems. We propose a Knowledge-Aware Two-tower Framework (KATT) that synergises Knowledge-Aware Networks (KAN) with a lightweight attention mechanism to achieve both high performance (Area Under the Curve (AUC): 0.8645, Accuracy (ACC): 78.13%) and interpretability. Evaluations on the MoocCube dataset demonstrate KATT’s superiority over six state-of-the-art baselines, particularly in cold-start scenarios (15% improvement) and recommendation explainability. The framework’s balanced approach to pedagogical relevance and system efficiency offers a novel solution for scalable personalised education.
Funding Information
  • Key Project of Zhejiang Provincial philosophy and Social Sciences Planning (25NDJC030Z)
  • 2022 Zhejiang Provincial Ideological and political Education Demonstration Grassroots Teaching Organization Project: Digital and Intelligent Teaching Innovation Team
  • The 2022 fourteenth five-Year Plan Educational Reform Project for Ordinary Undergraduate Universities in Zhejiang Province:Research and Practice of a Project-Led Software Competency Development System (1000416230660203)
  • Key Project of the Zhejiang Soft Science Research plan: Research on the construction and synergistic Pathways of Regional Innovation Ecosystems Oriented Towards New Forms of productivity (59)
  • Key Project of the Zhejiang Provincial Maior Humanities and Social Sciences Research Program (2024GH034)
  • The research Project of the Zhejiang University of Technology's Digital Innovation and Global value chain Upgrading Research Center

This publication has 16 references indexed in Scilit: