Attention-based deep neural network for Internet platform group users' dynamic identification and recommendation

Abstract Under the Internet background, group recommendation has become a major interest in the study of recommendation systems. In the method of group recommendation, the existing researches are mostly conducted using cluster analysis and similarity analysis. The group characteristics studied are relatively generalized, and the group objects studied are mostly fixed, so the group cannot change in real time according to the different attributes and characteristics of the different products. At the same time, for the research object of group recommendation, the existing research mainly consider the recommended project group or user group, but seldom consider recommending the appropriate project group to the appropriate user group to improve the recommendation efficiency and user satisfaction. In view of these problems, this paper proposes a deep neural network that integrates the attention mechanism for group users’ dynamic identification and recommendation on the Internet platform. This paper uses an attention mechanism and deep neural networks to generate the attention preference weights for the group users according to the product attributes. Doing so achieves the purpose of recommending many types of projects to different groups to adapt to their preferences. We compare this method with other baseline methods on two public datasets to validate the effectiveness of the proposed method, which achieves better performance than the most advanced methods.