Semantic overlapping community detection with embedding multi-dimensional relationships and spatial context

Semantic overlapping community detection is an important research hotspot in social networks. However, most of the existing methods ignore the fusion of multi-semantic attributes of users and variability of topological features under the spatial context. Motivated by this, we propose a semantic overlapping community detection method with embedding multi-dimensional relationships and spatial context. Firstly, on the basis of modeling users’ Microblogs by LDA, we further extract the semantic topic information. Two core topic matrices are captured to quantify users’ interest preferences. And the structure of social network is optimized by a new index named Network Constructability in line with user interest preference. Secondly, taking the network topology embedded with implicit semantic information as the core, the seed nodes are generated by leveraging Entropy to fuse explicit information. Finally, under the spatial context, Friend Probability and the Preference Matching Degree are utilized to deeply search the corresponding neighbor nodes in the process of detection. The effectiveness of the method is verified on the real datasets of social networks. The experimental results show that the method has obvious advantages and can detect the high-quality semantic overlapping communities with spatial context in real social networks.