Link Clustering with Extended Link Similarity and EQ Evaluation Division

Area of Science:

• Network Science
• Data Mining
• Computational Social Science

Background:

• Link Clustering (LC) is a community detection method using neighbor link similarity and hierarchical clustering.
• The original LC method has limitations, including not considering non-neighbor links and potentially creating too many small communities.

Purpose of the Study:

• To propose an Extended Link Clustering (ELC) method for improved overlapping community detection.
• To enhance the accuracy and sensibility of community detection in complex networks.

Main Methods:

• Introduced Extended Link Similarity (ELS) to create a denser transform matrix.
• Utilized the maximum Extended Quality of Modularity (EQ) value for optimal dendrogram cutting.
• Applied hierarchical clustering to the enhanced transform matrix.

Main Results:

• The ELC method achieved higher EQ and In-group Proportion (IGP) values compared to existing methods.
• Generated more realistic and sensible communities on real-world and artificial networks.
• Demonstrated superior performance over the original LC and classical CPM methods.

Conclusions:

• Extended Link Clustering (ELC) offers a more effective approach for overlapping community detection.
• The use of ELS and EQ optimization leads to improved network partitioning.
• ELC provides a robust framework for analyzing community structures in networks.