Health Discussions in Online Social Networks

This project is developing novel graph theory approaches to understand human networks (online social networks and COVID-19 Test, Trace and Isolate contact networks). It thus combines ICT and Mathematical Sciences and has a supervisor from each area.

The objectives of the project fall into three sub-projects:
Developing a novel distance measure: Given a set of complex networks, represented by the mathematical object graph, come up with a distance measure, such that graphs generated by the same mechanism are closer (distance is shorter) than graphs generated by different mechanisms.
Investigating how the structure and sentiment of vaccination discussions in online social networks (Twitter and Mumsnet) change over time, comparing discussion of vaccination pre-COVID with discussion post-COVID, to understand whether there are differences in how people consider this often contentious topic as a new vaccine is developed and deployed.
Modelling the impact of differences in the contact network structure on contact tracing strategies for COVID infections.

The contributions of the sub-project are as follows:
- Using a pre-existing distance measure, NetEmd, we adapt it to directed networks and improve its accuracy by combining the inputs with Principal Component Analysis and Independent Component Analysis.
- We compare how the structure of the social networks changes over time using NetEmd, determining what drives the change in structure and the effect of real-world events on discussions about vaccination. We apply pre-existing sentiment analysis techniques to these datasets and match changes in structure with changes in sentiment.
- Using the agent-based model of COVID simulation, covasim, we implement contact networks generated from age-mixing matrices, preferential attachment and contact data of the UK contact survey CoMix. We explore the effectiveness of contact tracing and testing strategies by varying the underlying structure of the contact networks