Discovery of User Groups Densely Connecting Virtual and Physical Worlds in Event-Based Social Networks

Abstract

An essential task of the event-based social network (EBSN) platform is to recommend events to user groups. Usually, users are more willing to participate in events and interest groups with their friends, forming a particularly closely connected user group. However, such groups do not explicitly exist in EBSN. Therefore, studying how to discover groups composed of users who frequently participate in events and interest groups in EBSN has essential theoretical and practical significance. This article proposes the problem of discovering maximum k fully connected user groups. To address this issue, this article designs and implements three algorithms: a search algorithm based on Max-miner (MMBS), a search algorithm based on two vectors (TVBS) and enumeration tree, and a divide-and-conquer parallel search algorithm (DCPS). The authors conducted experiments on real datasets. The comparison of experimental results of these three algorithms on datasets from different cities shows that the DCPS algorithm and TVBS algorithm significantly accelerate their computational time when the minimum support rate is low. The time consumption of DCPS algorithm can reach one tenth or even lower than that of MMBS algorithm.