Article Preview
Top1. Discovery Of User Groups Densely Connecting Virtual And Physical Worlds In Event-Based Social Networks
Event-Based Social Network (EBSN) is a new type of social network that can provide online social services and offline social activities (Liu et al., 2012). Online social services typically include users joining online interest groups, sharing comments and photos, etc. EBSN provides many offline social activities, such as reading, dancing, swimming, etc. The services provided by EBSN have attracted a large number of users and are increasingly favored by them. Typical EBSN platforms include Meetup, Plancast, Doubanevent, and more (Lan et al., 2022).
In EBSN, users usually participate in events and interest groups with their familiar friends. These people influence each other when deciding to participate in events or interest groups, forming a particular group. Usually, members of such groups significantly affect each other and are particularly closely connected, and they will participate in multiple interest groups and events together. Treating such groups as a whole for event recommendation would have a better effect than only considering individual user event recommendations. Kim et al. (2020) proposed that group users maintain close connections on social networking platforms and geographical locations, which can facilitate event recommendation, friend recommendation, and geographic data analysis. In addition to considering group members’ preference similarity, online and offline interactions between members can also be considered, utilizing various information to discover groups (Liao et al., 2021). Wang et al. (2017) believe that appropriate aggregation of online and offline user groups can better understand user behavior and its underlying organizational principles comprehensively. Trinh et al. (2020) defined group activity and user loyalty and proposed a method to measure group activity. The article believes that the close relationship between users and groups determines group activity, and user loyalty is a critical factor in maintaining group activity. The close online and offline connections between users and groups promote the development of the group. The above research indicates that identifying user groups with close online and offline relationships is meaningful and deserves our in-depth study.
The EBSN recommendation system effectively improves the efficiency of recommendation and prediction by recommending events and predicting group event participation for closely connected groups online and offline. However, EBSN does not directly provide such groups, and finding suitable and effective group discovery methods to quickly and accurately discover such groups is the purpose of this study.
In Figure 1, Mary, Tom, John, Bob, and Li jointly participated in multiple interest groups and events, forming a particularly closely connected group of users. However, they also participated in the event and interest groups together with Tony, and among these six user groups, Tony and other users were not closely connected, which is not the group we want to discover.
Figure 1. A closely connected group in EBSN
There are two main methods for obtaining groups in EBSN. One is to cluster users who participate in the same event or interest group or who have high similarity in EBSN. Another type is to divide communities based on user similarity.
Liao et al. (2019) removed inactive users from their interest groups and treated the remaining users as groups. Liao et al. (2006) formed groups of users who frequently participate in events together. Yuan et al. (2014) and Vinh et al. (2019) formed a group of users who participate in a certain event. Du et al. (2019) defined that users with preference similarity higher than a specific threshold form user groups. Jeong, et al. (2019), Ji et al. (2018), and Liao et al. (2020) identified users from the same interest group as user groups. Purushotham et al. (2016) proposed using offline social groups participating in events at specific locations as user groups. Trinh et al. (2021) searched for users and their friends and recommended events to them. The group obtained by this method may not have close connections among its members and may not necessarily participate in the next event together.