Multiobjective Optimization and Local Merge for Clustering Attributed Graphs.

Area of Science:

• Network Science
• Data Mining
• Computational Social Science

Background:

• Traditional community detection methods primarily rely on network topology.
• The importance of node attributes (compositional dimension) for accurate community structure is increasingly recognized.
• Existing methods often overlook the rich content information inherent in attributed networks.

Purpose of the Study:

• To propose a novel multiobjective genetic framework for community detection in attributed networks.
• To integrate both topological and compositional network dimensions for improved community discovery.
• To enhance the accuracy of community detection by leveraging node attributes and network structure.

Main Methods:

• Development of a multiobjective genetic framework.
• Integration of network topology measures for identifying densely connected communities.
• Utilization of attribute similarity measures for high intracommunity feature homogeneity.
• Implementation of an efficient post-processing local merge procedure.

Main Results:

• The proposed framework effectively uncovers community structure in attributed networks.
• Experimental results on synthetic and real-world networks demonstrate high-quality solutions.
• The approach outperforms several state-of-the-art community detection methods.
• Demonstrated efficacy in generating accurate network divisions reflecting group organization.

Conclusions:

• Integrating topological and compositional dimensions significantly improves community detection in attributed networks.
• The multiobjective genetic framework offers a robust and effective approach for uncovering complex community structures.
• The method provides a valuable tool for analyzing real-world networks with rich node attributes.