Evaluating the teaching quality level of university courses is a very meaningful teaching research work. In the context of big data, various artificial intelligence algorithms can effectively utilize massive data to construct intelligent models for evaluating the teaching quality of universities. Popular algorithms include genetic algorithm, BP (Back Propagation) neural network, and support vector machine. However, the modeling steps of these algorithms are usually cumbersome and time-consuming. In response to this, this article proposed a university teaching quality evaluation model based on the Naive Bayes classification algorithm. The model calculated the posterior probability of labels using Bayesian formula and the posterior probability of features by statistically analyzing the feature distribution in the sample. The label with the highest posterior probability was selected as the evaluation result for the feature vector. This article designed experiments to verify the performance advantages and disadvantages of this model compared to other models, and compared the modeling time, accuracy, mean square error, and disciplinary generality of the Naive Bayes model with the three aforementioned models. The results indicated that the modeling time of the Naive Bayes model was 0.12 seconds, with an accuracy of 0.895 and a mean square error of 0.1902. The accuracy difference between science and philosophy disciplines was 0.029, with a mean square error difference of 0.0190. Although the model in this article does not have an advantage in prediction accuracy compared to other models, it has good disciplinary universality, mainly due to its simple model structure and extremely short modeling time.