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

Area of Science:

• Data Mining and Network Analysis
• Social Network Analysis
• Geographic Information Systems

Background:

• Location-based social networks (LBSNs) integrate user location and social interactions.
• Community detection in LBSNs requires considering both spatial and social factors.
• Existing methods often struggle to effectively combine these dual factors for meaningful insights.

Purpose of the Study:

• To propose a novel geo-social group (GSG) model that unifies spatial and social information.
• To develop an efficient algorithm (GSGD) for detecting GSGs in large-scale LBSNs.
• To extend the problem to top-k geo-social group (TkGSG) mining for diverse community discovery.

Main Methods:

• Formulation of the Geo-Social Group (GSG) model defining group-venue clusters within a Minimum Covering Circle (MCC) threshold (γ).
• Development of a three-step GSGD algorithm for enumerating qualified GSGs with optimization techniques.
• Proposal of a greedy approach with a proven approximation ratio (1-1/e) for the NP-hard TkGSG problem.

Main Results:

• The proposed GSGD algorithm efficiently enumerates communities in large networks.
• The greedy approach for TkGSG mining provides a guaranteed approximation ratio.
• Empirical studies demonstrate the superiority of the proposed model and algorithms over existing methods.

Conclusions:

• The novel GSG model effectively integrates spatial and social factors for enhanced community detection.
• The developed algorithms offer efficient solutions for both full GSG detection and top-k mining.
• The research provides significant advancements in understanding and analyzing geo-social communities in LBSNs.