To improve the teaching level of saxophone performance in colleges and universities, this paper constructs a saxophone performance education optimization platform based on Bayesian theory. Bayesian clustering is used to informally cluster the saxophone performance education situation in colleges and universities, and the maximum likelihood algorithm is used to calculate the student behavior parameters. The Bayesian network tracking technique analyzes all learning behavior indicators to serve as a reference to obtain labels of learner motivation. To ensure the validity of the feedback metrics, the Bayesian stochastic calculation method was invoked to calculate the AUC (metric) mean ratio of the three feedback metrics. A practical teaching analysis was conducted to verify the platform’s timeliness. The results show that: The response time of login and resource search of the platform constructed in this paper is less than 2.5s and 1.2s, respectively, and the AUC mean ratio of three feedback metrics, namely, download rate, registration rate, and click rate, has improved by about 0.75%. Moreover, the positive attitude of university students towards learning has reached 88%. This shows the reliability of the educational platform constructed in this paper in optimizing higher education saxophone performance education, which can effectively improve the teaching level of saxophone performance education.