A multi-similarity spectral clustering method for community detection in dynamic networks

Area of Science:

• Network Science
• Data Mining
• Machine Learning

Background:

• Community structure is key in complex networks.
• Existing methods often fail with dynamic, evolving networks.
• Evolutionary clustering offers a framework for dynamic data and networks.

Purpose of the Study:

• To propose an improved evolutionary clustering method for dynamic network community detection.
• To address limitations of static network methods in evolving network scenarios.
• To enhance community detection accuracy in time-varying network structures.

Main Methods:

• Developed a multi-similarity spectral clustering (MSSC) method.
• Constructed multiple similarity matrices for each network snapshot.
• Employed a dynamic co-training algorithm using bootstrapped clustering across different similarity measures.

Main Results:

• The proposed MSSC method demonstrated superior performance compared to baseline models.
• Experiments were conducted on both synthetic and real-world dynamic network datasets.
• MSSC effectively identified community structures that change over time.

Conclusions:

• The multi-similarity spectral clustering method is effective for dynamic network community detection.
• Considering multiple similarity metrics enhances the performance of evolutionary clustering.
• MSSC provides a robust approach for analyzing evolving network structures.