This paper proposes a collaborative filtering-based learning resource recommendation algorithm to help learners find interesting resources in a huge amount of teaching resources and improve learning efficiency and learning interest so that the classical dance teaching content can be further enriched and expanded. By improving the K-means clustering algorithm, the article realizes the clustering of students’ classical dance teaching content preferences and further proposes a collaborative filtering algorithm based on ratings and attribute preferences of classical dance teaching resources, which takes into account the relationship between users’ needs and the attributes of the resources in order to complete the personalized recommendation of classical dance teaching resources for students. The article finally takes a dance teaching classroom as an example for empirical research. In the two groups of classical dance performance scoring a single total average score, the experimental group’s performance of dance movements is better than that of the control group, the average score of the experimental group’s pre-test is higher than the average score of the control group’s pre-test by 8±1.8 points, and the average score of the experimental group’s post-test is higher than the average score of the control group’s post-test by 9±2 points.