Abstract This paper discusses the implementation of informatization strategies in music teaching in colleges and universities, mainly through constructing an online teaching platform, using Bayesian nonparametric models and extreme gradient boosting algorithms to optimize the teaching process by classifying music and recommender systems. The study collected students’ specific information and learning styles, and found that most students solved problems by searching for tutorial books and consulting their classmates, and less often used online resources. The online teaching platform generated feedback reports based on the number of interactions, engagement and attendance. Classroom activities focused mainly on interaction and discussion in the first 15 minutes, while whole-class quizzes were conducted in the 18 to 30 and 40 to 50 minute segments. Finally, the teaching practice was evaluated through a questionnaire, and the results showed that the p-values of the five dimensions of aesthetic perception, artistic performance, cultural understanding, interactive learning and interest inquiry were significantly less than 0.001, indicating that the implementation of informative teaching strategies enormously helped to enhance students’ learning outcomes.