In recent years, the rapid development of artificial intelligence technology represented by knowledge graphs and deep learning has provided an opportunity for educational innovation and learning mode change. A smart music learning model for colleges and universities is developed in this paper with the help of artificial intelligence technology. Learning data analysis is achieved through the use of a community discovery algorithm based on graph data in the model. In order to construct the learning community, the AGNES hierarchical clustering algorithm is used to cluster individual samples in the dataset. Learning big data and music professional ability are correlated through the mining of learner portrait features. Personalized learning paths are generated using the improved convolutional neural network. As experimental subjects, sophomore music majors at X institution were tested and analyzed for the teaching model in the end. The results show that the maximum learning interaction coefficient of the experimental subjects can be obtained as 5.34 and the maximum learning path coefficient as 2.84 under the smart learning mode. The correlation coefficients of the use of the smart learning mode with the usual test scores and the learning effort values are between 0.318 and 0.502. Teachers can obtain precise teaching data from this paper to quantitatively characterize subject competence goals and facilitate the smooth implementation of smart learning.