Abstract:With the development of mobile Internet, Event-based Social Networks (EBSN) have been born and flourished. Users in Event-based Social Networks can not only interact online through the services provided by online social platforms, but also participate in social activities face to face. This paper studies the problem of personalized social event recommendation in Event-based Social Networks, and proposes a personalized social event recommendation method which integrates user's historical behavior and social relationship. At present, the methods of personalized event recommendation in Event-based Social Networks mainly adopt the methods based on graph model and influencing factors. When dealing with large-scale social networks and data sparsity problems, the methods based on graph model are inefficient and ineffective. However, when modeling the entities in the network, the method based on influencing factors usually considers the time and place of the event rather than the relevant factors of users, and ignores the important role of users in the social network. In this paper, deep learning method is used to study the personalized social event recommendation problem in Event-based Social Networks. User profiles are built from two aspects of user's historical behavior and potential social relationship. In user modeling, negative vector representation of users is introduced, and attention mechanism is used to recommend different events for different historical behaviors of users according to different candidate events and assign different weights to different friends in the social relationship, and consider various characteristics of events and groups. At last, we do a lot of experiments on the real world dataset. The experimental results show that the personalized social event recommendation method proposed in this paper is better than the current algorithm in many evaluation indexes.