Self-focus Sequence Recommendation Model for Fusion Context Information

Abstract

Sequence recommendation is a hot research content in the system. The core idea of sequence is to dig all the relationships between users and projects from the sequence of users and projects, and personalize the user's next item may interact. Existing Most of the research methods are modeling the user ID and the project ID interactive sequence, and the impact of the project and user context information is ignored. For this issue, this paper proposes a converged project and user context information. Self-focus sequence recommendation model. This model consists of two parallel modules of the embedded layer module, convolutional neural network and self-focus mechanism network, where the convolutional neural network module is modeled for dynamic preferences of the user and the project interaction. Self-focus mechanism network module capture user and user contextual information characteristics. Finally, the user's dynamic preferences and projects learned from the two modules are combined with the user context information feature to enhance the recommended performance. The experimental results show that the model of this paper has increased by at least 23.2%, the accuracy rate and the recall rate are also significantly improved compared to the current baseline method.