Network Sampling Techniques in Social Graphs

Network sampling in social graphs encompasses a suite of methodologies designed to extract representative subsets from large‐scale networks, enabling analysis when complete data are unavailable or impractical to collect. Core approaches include node‐based sampling, edge‐based sampling and traversal‐based sampling, each aiming to preserve key topological features such as degree distributions, clustering coefficients and community structures. Random walk sampling and its variants, including the Metropolis–Hastings random walk, are widely used to mitigate selection bias by probabilistically visiting nodes according to predefined transition rules. Snowball and respondent‐driven sampling techniques explore local neighbourhoods to capture relational patterns, whereas stratified or attribute‐driven schemes select nodes or edges based on observable characteristics. Advanced methods address the core–periphery dichotomy by adapting sampling rates to node centrality, thereby balancing the focus on densely connected hubs and sparsely linked peripheral nodes. Recent developments have emphasised stochastic modelling of link weights, Bayesian reconstruction of global statistics from partial observations and double‐layer strategies that distinguish between high‐ and low‐degree regions. These innovations facilitate applications ranging from influence maximisation and information diffusion to epidemic modelling and marketing analysis, underlining the global significance of robust sampling in understanding and harnessing social connectivity.

Research from Nature Portfolio

Recent studies have introduced a diffusion framework that treats influence probabilities on social ties as random variables, estimating these through iterative sampling and adaptive learning automata. This stochastic model offers a flexible alternative to deterministic diffusion, demonstrating improved performance in influence maximisation tasks on both real and synthetic networks. Another advancement addresses sampling of unknown large‐scale networks at very low sampling rates. By first estimating a degree threshold to partition the network into core and periphery, a double‐layer sampling strategy is applied that preserves crucial structural properties with high efficiency. Experimental results confirm that this method accurately retains degree distributions and clustering patterns in scale‐free topologies, even when only a small fraction of nodes is observed.

Network Sampling Techniques in Social Graphs publication trend

The graph below shows the total number of articles in network sampling techniques in social graphs across all publications each year (not limited to Nature Index journals).

Technical terms

Random walk sampling: A traversal method that selects nodes by following edges according to a probability distribution, reducing selection bias.

Metropolis–Hastings random walk: An extension of random walk sampling that adjusts transition probabilities to achieve a desired stationary distribution.

Edge sampling: Technique in which edges are selected uniformly or with bias, yielding a subgraph from which global statistics can be inferred.

Stochastic graph: A network model in which link weights or existence are treated as random variables, capturing uncertainty in connections.

Core–periphery structure: Division of a network into densely connected core nodes and sparsely connected peripheral nodes, guiding targeted sampling.

Influence maximisation: The problem of identifying a set of nodes whose activation yields the largest expected spread of information or behaviour.

Bayesian reconstruction: Statistical method that leverages prior distributions to infer network properties from sampled observations.

References

1. Towards a standard sampling methodology on online social networks: collecting global trends on Twitter. Applied Network Science (2016).
2. A new stochastic diffusion model for influence maximization in social networks. Scientific Reports (2023).
3. Benefits of Bias in Crawl-Based Network Sampling for Identifying Key Node Set. IEEE Access (2020).
4. Sampling unknown large networks restricted by low sampling rates. Scientific Reports (2024).
5. Using a Bayesian approach to reconstruct graph statistics after edge sampling. Applied Network Science (2023).