Construction of a Vocational College Ecological Course Recommendation System Using BERT and Knowledge Graph

With the acceleration of digitalization in vocational education, the curriculum system for ecological majors has become increasingly complex. Students often face information overload and low matching accuracy when selecting courses. Traditional recommendation systems mainly rely on collaborative filtering or simple content features, making it difficult to capture deep semantics in course texts and knowledge connections among courses. To address this issue, this paper proposes a vocational college ecological course recommendation system integrating BERT and a knowledge graph. First, BERT is used to semantically encode course text information, including syllabi and teaching objectives, to generate high-precision course feature vectors. Second, an ecological course knowledge graph covering courses, knowledge points, professional requirements, and student portraits is constructed to mine multidimensional relationships. Finally, a hybrid recommendation algorithm combining semantic similarity and knowledge graph correlation is designed to achieve personalized course recommendation. Experimental results show that the system outperforms traditional recommendation methods in precision, recall, and F1-score, providing more accurate and instructive course recommendation services for ecological majors in vocational colleges.