Link community detection using generative model and nonnegative matrix factorization.

Area of Science:

• Network Science
• Data Mining
• Computational Biology

Background:

• Community discovery is crucial for understanding complex networks.
• Existing methods primarily focus on node communities, overlooking link communities.
• Generative models offer potential for modular structure identification but often neglect link communities.

Purpose of the Study:

• To propose a novel generative model for link community discovery.
• To address the limitations of existing models in detecting link communities.
• To develop an efficient and automated method for identifying link communities in large networks.

Main Methods:

• A generative model based on node importance within communities was developed.
• Non-negative matrix factorization was employed for parameter fitting.
• An iterative bipartition strategy was introduced for automatic community number determination.

Main Results:

• The proposed model successfully describes link community structures.
• The iterative bipartition strategy automatically determined the number of communities.
• The approach demonstrated superior performance compared to related methods on synthetic and real-world networks, including a biological network.

Conclusions:

• The developed generative model offers an effective approach for link community detection.
• The method's efficiency and automated nature make it suitable for large, unexplored networks.
• This work advances the field of community discovery by focusing on link communities.