DSH2025

Modeling and Mining High-order Interactions in Social Media Data.

Copenhagen, 23 June 2025

13:15 - 17:00

Room 2.1.042 (AAU)

About the tutorial

Social media platforms generate vast amounts of data characterized by complex, multi-faceted interactions that extend beyond simple pairwise relationships. While traditional graph based approaches have been widely used to analyze these interactions, they fail to capture inherent group dynamics.
This tutorial introduces hypergraphs as a robust mathematical framework for modeling and analyzing high-order interactions in social media data. Hypergraphs, which allow (hyper)edges to connect an arbitrary number of nodes, can naturally represent group interactions such as multi-user conversations, shared content engagement, and community behaviors. This tutorial covers both theoretical foundations and practical implementations, with a special focus on comparing hypergraph-based analyses with traditional graph-based approaches.
Through hands-on exercises using Python and Colab Notebooks, participants will learn how to build hypergraphs from social media data, compute fundamental metrics, detect communities, and deal with more advanced topics, such as analyzing information diffusion patterns. Participants will work with real social media datasets to understand the advantages and challenges of hypergraph modeling. By the end of the tutorial, attendees will have gained practical experience in applying hypergraph analysis to social media data and understanding when and how to leverage this powerful analytical framework.

Schedule

Topics covered

Part I: Foundation & data preparation

Introduction to hypergraphs and social media data
Setting up the programming environment
Loading and preprocessing sample social media data
Exercise: Building hypegraphs from sample social media data

Part II: Analysis techniques

Basic hypergraph metrics (node centrality, hyperedge size distribution)
Community detection in hypergraphs
Visualization techniques for hypergraphs
Exercise: Identifying influential nodes and important groups through hypergraph centrality measures

Part III: Advanced applications

Information diffusion patterns in hypergraphs
Exercise 1: Track the spread of a topic in a social media network modeled as a hypergraph
Compare the results of the previous exercise with the ones obtained by modeling the network with a graph

Materials

All materials will be available on our website and a publicly available GitHub repository beforehand. These include pre- prepared Colab/Jupyter notebooks, Python scripts, library settings, and slides. We will also use publicly available real- world networks.

Additional information

Participants are only required to have a browser installed on their own device and a Google account to run the Co- lab Notebooks we will make available. No specific software licenses are required, and the setup should be almost imme- diate.

Target

This tutorial targets researchers interested in analyzing high- order relations in social media through the lens of hyper- graph theory and analysis. While programming experience is beneficial, it is not mandatory.