Fast generation of simple directed social network graphs with reciprocal edges and high clustering

TL;DR

This paper presents a fast, scalable method for generating synthetic directed social network graphs with properties similar to real-world networks like Twitter, focusing on reciprocal edges and high clustering.

Contribution

It introduces a new graph generation approach based on a previous model, with fewer hyperparameters and lower runtime, improving scalability and fidelity.

Findings

The method accurately replicates Twitter-like topological features.

It effectively models epidemic spreading processes on generated graphs.

The approach is highly scalable for larger network simulations.

Abstract

Online social networks have emerged as useful tools to communicate or share information and news on a daily basis. One of the most popular networks is Twitter, where users connect to each other via directed follower relationships. Researchers have studied Twitter follower graphs and described them with various topological features. Collecting Twitter data, especially crawling the followers of users, is a tedious and time-consuming process and the data needs to be treated carefully due to its sensitive nature, containing personal user information. We therefore aim at the fast generation of synthetic directed social network graphs with reciprocal edges and high clustering. Our proposed method is based on a previously developed model, but relies on less hyperparameters and has a significantly lower runtime. Results show that the method does not only replicate the crawled directed Twitter graphs well w.r.t. several topological features and the application of an epidemics spreading process, but that it is also highly scalable which allows the fast creation of bigger graphs that exhibit similar properties as real-world networks.