Model-based clustering of time-evolving networks through temporal exponential-family random graph models

Area of Science:

• Network Science
• Statistical Modeling
• Data Analysis

Background:

• Dynamic networks model complex systems evolving over time.
• Detecting groups with similar connectivity in these networks is a key challenge.
• Existing methods may not fully capture temporal dynamics or group structures.

Purpose of the Study:

• To develop a model-based clustering framework for time-evolving networks.
• To simultaneously model network structure and detect group patterns.
• To provide an effective criterion for selecting the optimal number of groups.

Main Methods:

• Utilized discrete time exponential-family random graph models.
• Developed a conditional likelihood approach for model selection.
• Implemented an efficient variational expectation-maximization (EM) algorithm for parameter estimation.

Main Results:

• The proposed framework effectively models and detects group structures in dynamic networks.
• The model selection criterion accurately determines the number of groups.
• The EM algorithm provides efficient parameter and mixing proportion estimation.

Conclusions:

• The developed framework offers a robust approach for analyzing group structures in dynamic networks.
• Demonstrated applicability in simulations and real-world networks like international trade and academic collaborations.
• Provides a valuable tool for understanding evolving complex systems.