This paper uses the BTM algorithm, which has been improved by word embedding and self-attention mechanisms, to analyze the degree of embedding of innovation theory in college students' management. By analyzing the principle of attention mechanism and self-attention mechanism, it constructs the embedding analysis model for higher education innovation theory. By analyzing the cw2vec and word2vec models, the expression of innovation theory vocabulary is derived. Based on the improved BTM algorithm, the WESA-BTM model is applied to give higher weights to the subject words with higher subject contributions. Parameter solving, as well as derivation, were used to determine the impact of innovation theory on college students’ management follow-up. The results show that the improved BTM algorithm can effectively analyze the embeddedness of innovation theory in college students’ management, and the WESA-BTM model in Web Snippets is evaluated to be -573.26 at the number of topics of 20 and -561.71 at the number of topics of 40. This study improves the efficiency of college students’ management in colleges and universities to a certain extent. © 2023 Chunlin Yang, published by Sciendo.
