[Paper Review] The Anatomy of the Facebook Social Graph

Johan Ugander, Brian Karrer+2 more|arXiv (Cornell University)|Nov 18, 2011

Complex Network Analysis TechniquesPhysics and Astronomy49 references781 citations

TL;DR

This study presents the largest-scale analysis of a social network to date, using anonymized data from 721 million active Facebook users in May 2011. It reveals that the Facebook social graph is nearly fully connected (99.91% in a single giant component), exhibits a small-world structure with an average path length of 4.7, and displays dense local neighborhoods despite global sparsity, while also identifying strong community structure driven by nationality and age homophily.

ABSTRACT

We study the structure of the social graph of active Facebook users, the largest social network ever analyzed. We compute numerous features of the graph including the number of users and friendships, the degree distribution, path lengths, clustering, and mixing patterns. Our results center around three main observations. First, we characterize the global structure of the graph, determining that the social network is nearly fully connected, with 99.91% of individuals belonging to a single large connected component, and we confirm the "six degrees of separation" phenomenon on a global scale. Second, by studying the average local clustering coefficient and degeneracy of graph neighborhoods, we show that while the Facebook graph as a whole is clearly sparse, the graph neighborhoods of users contain surprisingly dense structure. Third, we characterize the assortativity patterns present in the graph by studying the basic demographic and network properties of users. We observe clear degree assortativity and characterize the extent to which "your friends have more friends than you". Furthermore, we observe a strong effect of age on friendship preferences as well as a globally modular community structure driven by nationality, but we do not find any strong gender homophily. We compare our results with those from smaller social networks and find mostly, but not entirely, agreement on common structural network characteristics.

Motivation & Objective

• To characterize the global structural properties of the world’s largest social network, Facebook, at scale.
• To investigate the extent of connectivity, path lengths, clustering, and community structure in the Facebook social graph.
• To examine demographic mixing patterns, including homophily by age, gender, and country, and their impact on network structure.
• To compare findings with smaller social networks and assess the generalizability of known network properties at massive scale.

Proposed method

• Analyzed the entire Facebook social graph of active users (n ≈ 721 million) using anonymized data from May 2011.
• Employed the Newman-Zipf (NZ) algorithm to compute connected component structure via streaming edge processing on a single machine.
• Used the HyperANF algorithm on a 24-core machine to estimate average shortest path lengths across 10 runs.
• Applied reservoir sampling to select 500,000 users (5,000 per 100 log-spaced neighborhood sizes) for local network analysis.
• Computed empirical percentiles of clustering coefficient and degeneracy across sampled user neighborhoods.
• Leveraged Hadoop/Hive on a 2,250-machine cluster for large-scale data processing and network feature extraction.

Experimental results

Research questions

• RQ1 To what extent is the Facebook social graph globally connected, and what is the average path length between users?
• RQ2 How do local network neighborhoods of users compare in density and clustering to the global graph structure?
• RQ3 What role do demographic factors such as age, gender, and nationality play in shaping friendship patterns and network mixing?
• RQ4 How does the community structure of the Facebook network manifest at the country and regional levels?
• RQ5 To what extent do network properties like degree assortativity and the 'friendship paradox' hold at scale in a real-world social network?

Key findings

• The Facebook social graph is nearly fully connected, with 99.91% of active users belonging to a single giant connected component.
• The average shortest path length between users in the giant component is 4.7, confirming the 'six degrees of separation' phenomenon on a global scale.
• Despite global sparsity, local neighborhoods of users exhibit high clustering and degeneracy, indicating dense structural cores around individuals.
• There is strong degree assortativity, and users consistently have friends who have more friends than they do, confirming the friendship paradox at scale.
• Age homophily is pronounced, with users showing a marked preference for friends of similar age, and nationality is a dominant driver of community structure.
• Geographical distance strongly influences inter-country friendships, with community structure at the country level largely shaped by proximity and historical/cultural ties.