Research On Personalized Recommendation Algorithm Of Learning Resources Based On Knowledge Graph

The rapid development of online learning platforms has brought about a huge number of online learning resources,while the massive amount of learning resources has also brought about the problem of information disorientation,making it difficult for students to quickly and accurately find the resources that match their learning abilities.In recent years,one of the research hotspots in the field of smart education is how to recommend high-quality personalized learning resources for students,so as to help smart education realize the concept of personalized teaching and learning according to their abilities.However,there are two major challenges in the field of learning resource recommendation: on the one hand,students’ learning behaviors take a longer time,which causes a more serious data sparsity problem in the field of learning resource recommendation compared with other traditional recommendation fields;on the other hand,students’ learning intentions are diverse and their interests change,so how to capture students’ dynamic preferences becomes another major challenge in the field of learning resource recommendation.In order to solve the above two problems,this paper proposes the learning resource recommendation model in this paper by studying and improving the classical models in the recommendation domain.The main work of this paper is as follows:(1)A graph convolution-based learning resource recommendation algorithm is proposed to address the problem of data sparsity.The algorithm first constructs a student-learning resource bipartite graph and captures students’ learning preferences and similar learning resources through lightweight graph convolution.Subsequently,the learning resources in the student-learning resource bipartite graph are bridged to the same learning resource entities in the educational knowledge graph through entity alignment,and the graph convolution with attention mechanism is combined to explore the association of different types of learning resources and further enhance the item representation of learning resources,and finally the effectiveness of the algorithm is verified on the publicly available dataset MOOCCube.(2)An improved learning resource recommendation algorithm that fuses longand short-term dynamic preferences is proposed to address the problem that students’ complex learning intentions and dynamically changing learning interests are difficult to capture.The algorithm first takes short-term preferences,long-term preferences and context-influenced dynamic student representations as inputs,and then encodes them into long-term dynamic preference representations and short-term dynamic preference representations,respectively,and next aggregates the aforementioned student preference expressions captured by lightweight graph convolution as the final student interest embedding representations,and finally combines the aforementioned learning resource embedding representations with item-enhanced representations for student recommendation prediction,and likewise The effectiveness of the algorithm is verified on the MOOCCube dataset.