Sampling unknown large networks restricted by low sampling rates.

Area of Science:

• Data Mining
• Network Analysis
• Graph Theory

Background:

• Graph sampling is crucial for large network data mining.
• Low sampling rates in large networks bias traditional methods towards core nodes.
• Existing traversal-based sampling struggles with scale-free networks.

Purpose of the Study:

• To propose a novel sampling method (SLSR) for unknown large networks at low sampling rates.
• To address the bias of traditional methods towards densely-connected nodes.
• To accurately preserve critical network structures.

Main Methods:

• SLSR employs random node sampling to determine a degree threshold.
• It distinguishes between core and periphery network regions.
• A double-layer sampling strategy is applied to both core and periphery.

Main Results:

• SLSR demonstrates high time efficiency due to its simplicity.
• Experiments confirm accurate preservation of critical structures in scale-free networks.
• The method achieves low variances even with low sampling rates.

Conclusions:

• SLSR offers an effective solution for sampling large, unknown networks at low rates.
• The method overcomes the limitations of traversal-based sampling.
• SLSR provides a robust and efficient approach for network data mining.