With the continuous exploration of AI application fields, the application of “AI+Education” in the higher education market is also deepened. The article studies the personalized recommendation learning method based on deep knowledge tracking. The method adopts a dynamic key-value memory network that integrates the attention mechanism, and extracts and stores the features between students and marketing training exercises. Students’ dynamically changing marketing competency levels are tracked by a bidirectional gated recurrent unit neural network. The method of two exercises filtering is designed to calculate the cognitive similarity, the similarity of the difficulty of the exercises and the difficulty of the exercises of the students to achieve the recommendation of personalized exercises for senior marketing majors. The accuracy, novelty and diversity of the recommended exercises of this paper’s method are improved compared with the comparison model, and its recommendation accuracy is improved by 4.00% on the ASSISTment2009 dataset compared with the sub-optimal model. The posttest scores of the experimental class based on this paper’s personalized recommendation learning method gained significant improvement compared with the pre-test, and also gained significant improvement compared with the posttest scores of the control class. It shows that the method of this paper can provide reliable learning resources recommendation for senior marketing students and is suitable for personalized teaching by senior marketing teachers.
