Construction and Application of Personalized Learning Resources Based on Knowledge Graph in Higher Education Self-study Examination Environment

Abstract

 This paper first outlines the overall construction process of a course resource knowledge graph, analyzes the acquired dataset, and performs data cleaning and preprocessing. Relationships and entities are extracted from the dataset, and the course knowledge graph is constructed as triples and stored in the Neo4j graph database. Subsequently, semantic and difficulty information is extracted from exercise texts and learning records to obtain a comprehensive representation of students' responses at a given time. During the knowledge state update phase, unidirectional and bidirectional propagation modes are adopted for different relationships. Finally, against the backdrop of personalized learning in online education, a bipartite graph-based collaborative filtering course recommendation algorithm (CFBGR) is proposed, effectively mining the deep interaction information between learners and courses. The results show that the AUC and ACC values of the proposed model outperform those of some existing models, indicating that the research approach of considering knowledge point embedding in the knowledge tracking model has certain research significance. When recommending questions to students, the proposed method can tailor recommendations to each student's personalized learning level, ensuring that the recommended questions align with their individual needs. Additionally, the question recommendation results exhibit a high degree of accuracy and interpretability.