Measurement-calibrated graph models for social network experiments

Abstract

Access to realistic, complex graph datasets is critical to research on social networking systems and applications. Simulations on graph data provide critical evaluation of new systems and applications ranging from community detection to spam filtering and social web search. Due to the high time and resource costs of gathering real graph datasets through direct measurements, researchers are anonymizing and sharing a small number of valuable datasets with the community. However, performing experiments using shared real datasets faces three key disadvantages: concerns that graphs can be de-anonymized to reveal private information, increasing costs of distributing large datasets, and that a small number of available social graphs limits the statistical confidence in the results.The use of measurement-calibrated graph models is an attractive alternative to sharing datasets. Researchers can "fit" a graph model to a real social graph, extract a set of model parameters, and use them to generate multiple synthetic graphs statistically similar to the original graph. While numerous graph models have been proposed, it is unclear if they can produce synthetic graphs that accurately match the properties of the original graphs. In this paper, we explore the feasibility of measurement-calibrated synthetic graphs using six popular graph models and a variety of real social graphs gathered from the Facebook social network ranging from 30,000 to 3 million edges. We find that two models consistently produce synthetic graphs with common graph metric values similar to those of the original graphs. However, only one produces high fidelity results in our application-level benchmarks. While this shows that graph models can produce realistic synthetic graphs, it also highlights the fact that current graph metrics remain incomplete, and some applications expose graph properties that do not map to existing metrics.

LBSNSim: Analyzing and modeling location-based social networks

“…The CCDF is approximately a straight line in the log-log scale, which illustrates that user degrees follow a power law distribution. This result is consistent with the previous findings on online social networks [12], [16]: the majority of users have small degrees, while a small number of users have significantly larger degrees, which are the "hub" nodes in the social graph.…”

“…Sala et al explored the feasibility of replacing real social graphs of online social networks with synthetic graphs generated from calibrated graph models [16]. The authors compared six existing graph models.…”

“…Third, LBSN datasets with different properties can be generated on demand, which can help researchers improve the statistical confidence in their experimental results. Previous work investigated the graph models that produce synthetic social graphs of online social networks [8], [9], [16], [17]. Given all these advantages of the model-generated LBSN datasets, however, no LBSN model has been proposed in the literature.…”

Practical Conflict Graphs in the Wild

“…When calibrated using on-site measurements, however, these propagation models offer higher accuracy and have been used in cell planning [26], [27], interference management [33], and coverage prediction [28], [29]. Our work complements these prior works and is also inspired by prior work on measurement-calibrated models for social network graphs [54].…”

Towards Inferring Communication Patterns in Online Social Networks

“…These properties have in turn informed social network graphs generators that attempt to generate synthetic yet realistic social network graphs Sala et al (2010). Still, these models are limited in that they only describe the topology of the network and do not capture users' activity, this is, how users communicate and the actions they perform in the OSN.…”