A network community structure similarity index for weighted networks

Area of Science:

• Network analysis and complex systems science.
• Graph theory and data mining.

Background:

• Identifying communities in complex systems is crucial for network analysis.
• Existing methods for comparing community structures often ignore edge weights and node count differences.
• There is a need for robust similarity measures that account for these factors.

Purpose of the Study:

• To propose a novel network community structure similarity index (NCSSI).
• To address limitations of existing methods by incorporating edge weights and varying node counts.
• To provide a more comprehensive measure for comparing community structures across networks.

Main Methods:

• Developed a new similarity index, NCSSI, based on the edit distance concept.
• NCSSI integrates both community labels and edge weights for comparison.
• The method was validated using simulated data and a real-world case study (New York Yellow Taxi data).

Main Results:

• NCSSI effectively captures the impact of changes in both community labels and edge weights.
• The index demonstrates superior performance compared to traditional methods like mutual information and Jaccard index.
• NCSSI successfully handles comparisons between communities with differing numbers of nodes.

Conclusions:

• NCSSI offers a novel and effective approach for measuring community similarity in complex networks.
• The index provides a more nuanced understanding of community structure evolution and comparison.
• This method enhances the analysis of related networks with varying characteristics.