Significant Geo-Social Group Discovery over Location-Based Social Network

Wei Li,Sisi Zlatanova

doi:10.3390/s21134551

Wei Li, Sisi Zlatanova

Open Access

https://doi.org/10.3390/s21134551

Copy DOI

Abstract

Geo-social community detection over location-based social networks combining both location and social factors to generate useful computational results has attracted increasing interest from both industrial and academic communities. In this paper, we formulate a novel community model, termed geo-social group (GSG), to enforce both spatial and social factors to generate significant computational patterns and to investigate the problem of community detection over location-based social networks. Specifically, GSG detection aims to extract all group-venue clusters, where users are similar to each other in the same group and they are located in a minimum covering circle (MCC) for which the radius is no greater than a distance threshold . Then, we present a GSGD algorithm following a three-step paradigm to enumerate all qualified GSGs in a large network. We propose effective optimization techniques to efficiently enumerate all communities in a network. Furthermore, we extend a significant GSG detection problem to top-k geo-social group (TkGSG) mining. Rather than extracting all qualified GSGs in a network, TkGSG aims to return k feasibility groups to guarantee the diversity. We prove the hardness of computing the TkGSGs. Nevertheless, we propose the effective greedy approach with a guaranteed approximation ratio of . Extensive empirical studies on real and synthetic networks show the superiority of our algorithm when compared with existing methods and demonstrate the effectiveness of our new community model and the efficiency of our optimization techniques.

Highlights

Graphs have been widely used to model entities and their relationships
Different from the detection problem, the prerequisite of a community search requires an additional set of query vertices q, and the aim is to search for a set of cohesive subgraphs, each of which should contain all vertices of q
We formulated a problem of detecting significant geo-social groups (GSGs) over a massive graph with location labels on vertices, where a GSG has a cohesive social and spatial compactness

Summary

Introduction

Graphs have been widely used to model entities (as vertices) and their relationships (as edges). Most of the existing works on community detection and community search only consider the topological structure of a graph. With the proliferation of a network along with location such as Twitter, MeetUp, and Foursquare, interest on the topic of community detection/search over a geo-social network has increased [5,6,7,8,9]. In such networks, users are usually associated with location data, which is important for making sense of detected communities (e.g., checkins and hometown). A typical instance of a location-based social network is shown in

Objectives

Methods

Findings

Conclusion