Abstract: With the explosion of network information, more and more information is dizzying and the recommendation system came into being. This paper presents a recommendation system based on mutual information, which analyzes the data set from three aspects, including user behavior, product titles and product tags, using mutual information to calculate the similarity between items, verify the different weights, observe the effect on the recommended effect, and finally get the best recommended list. In addition, under the data set of this paper, the proposed method and the collaborative filtering recommendation are compared to prove that the proposed method is better than collaborative filtering.