Exploring Temporal Community Structure via Network Embedding.

Area of Science:

• Computer Science
• Network Science
• Data Mining

Background:

• Temporal community detection is crucial for analyzing dynamic networks.
• Existing methods struggle to capture network dynamics and community structures simultaneously.
• Network embedding and deep learning offer potential but require adaptation for temporal data.

Purpose of the Study:

• To propose a novel unsupervised dynamic community detection model.
• To effectively discover temporal communities and model dynamic networks.
• To improve the accuracy of community detection in evolving networks.

Main Methods:

• Developed an unsupervised dynamic community detection model based on network embedding.
• Introduced a community prior using Gaussian Mixture Model (GMM) within a variational autoencoder.
• Utilized a variant of Gated Recurrent Unit (GRU) to model evolutionary characteristics of community structure and node embedding.

Main Results:

• The proposed model effectively discovers temporal communities.
• It accurately models the evolutionary characteristics of dynamic networks.
• Extensive experiments show improved accuracy in dynamic community detection compared to existing methods.

Conclusions:

• The novel model successfully addresses limitations of previous approaches for temporal community detection.
• Integrating GMM and GRU with network embedding provides a powerful framework for dynamic network analysis.
• The method demonstrates superior performance on both real-world and artificial dynamic networks.