Comparing methods for comparing networks.

Area of Science:

• Network science
• Data analysis
• Computational methods

Background:

• The increasing volume of data necessitates robust quantitative methods for network comparison.
• Existing methods often focus on simple network types, limiting their applicability to complex, real-world networks.

Purpose of the Study:

• To review, classify, and compare existing network comparison methods.
• To provide a guide for selecting appropriate quantitative methods for network analysis.
• To assess the usability and meaningfulness of different network comparison techniques.

Main Methods:

• Classification of network comparison methods based on criteria, advantages, and drawbacks.
• Testing methods on synthetic networks to evaluate performance and result interpretability.
• Application of selected methods to real-world datasets (European Air Transportation Network, FAO Trade Network).

Main Results:

• Evaluation of methods requiring known node-correspondence (e.g., DeltaCon, Cut Distance) and those without (e.g., alignment-based, graphlet-based, spectral methods, Portrait Divergence, NetLSD).
• Assessment of method usability and the significance of their outputs on synthetic and real-world data.
• Comparative analysis of network comparison results across different datasets.

Conclusions:

• The study provides a comprehensive overview and practical guide to network comparison methodologies.
• Findings highlight the strengths and weaknesses of various methods for different network types and data complexities.
• The research aids researchers in selecting the most suitable quantitative methods for their specific network analysis tasks.