Community discovering in temporal network with spectral fusion

Area of Science:

• Network Science
• Data Mining
• Graph Theory

Background:

• Temporal networks model dynamic systems, but community detection remains challenging.
• Existing methods struggle with the time-varying nature and mutations of temporal networks.

Purpose of the Study:

• To develop a robust method for detecting invariable communities in temporal networks.
• To address the limitations of current community detection approaches in dynamic network analysis.

Main Methods:

• Proposed a temporal spectral clustering method.
• Integrated Fiedler's eigenvectors of normalized Laplacian matrices within a sliding time window.
• Incorporated historical network structure information.

Main Results:

• The proposed method effectively detects invariable communities in temporal networks.
• It outperforms existing methods, avoiding inaccurate partitions caused by network mutations.
• Experimental results validate the effectiveness of considering historical network information.

Conclusions:

• Temporal spectral clustering offers a promising approach for community detection in dynamic networks.
• Integrating historical structural information enhances the robustness and accuracy of community detection.
• The developed method provides a valuable tool for analyzing evolving network systems.