Abstract: The TF-IDF literature recommendation method based on the behavior data of electronic resources is proposed, aiming to improve the efficiency of using library point resources and realizing personalized and accurate literature recommendation services. Firstly, a corpus is formed by using the abstracts downloaded and browsed by users with the same professional background. Secondly, the abstracts in the corpus are divided into words and the TF-IDF values of the divided words are calculated. Then, the high-impact journal literature of the profession is selected as the literature to be recommended. Finally, the cosine similarity between the literature to be recommended and the TF-IDF values of each entry in the corpus is calculated, and the literature with high similarity is sent to the user. The empirical results show that, on one hand, the recommendation method is more accurate for users with similar academic research backgrounds, and, on the other hand, it provides new ideas for scholars to effectively recommend papers for research.