###Modelling infectious disease prevalence with AI methods and online user activity
Project ID: 2228bd1193 (You will need this ID for your application)
Research Theme: Healthcare Technologies
UCL Lead department: Computer Science
Lead Supervisor: Vasileios Lampos
Project Summary:
Previous research has showcased that online user activity is indicative of various health-related signals. In particular, Web search and social media trends can provide timely insights about epidemics of infectious diseases such as influenza or COVID-19. My group has been sharing insights about influenza and COVID-19 based on Google search activity with UK’s Health Security Agency (UKHSA) for more than 5 years.
This project will use alternative streams of information (Web search, social media, mobility patterns) to achieve all or a subset of the following aims. We are also open to other related research ideas within the theme of this project.
Develop machine learning (ML) methods for forecasting the prevalence of infectious diseases using online user-generated data.
Develop methods that can capitalise on the fine geographical granularity of online user activity data to provide early warnings for upcoming pandemics or seasonal epidemics as well as retrospectively understand how an infectious disease has spread.
Conduct research on the development of methodological frameworks that incorporate previously established approaches in epidemiology / biostatistics, e.g. compartmental / mechanistic models. Modern techniques based on neural network architectures provide superior accuracy but have no understanding of common disease transmission properties. What kind of model could bring the best of both approaches under the same optimisation task?
Develop open source, end-to-end solutions (e.g. a web service / dashboard) so that outcomes of our research can have a tangible impact by being shared with public health organisations.
We collaborate with UKHSA, WHO, Microsoft Research, and the EPSRC project i-sense. We have also support from Google that provides us with search activity trends.
The Ph.D. candidate should have a good understanding of introductory ML theory, good understanding of basic natural language processing (NLP) concepts, strong programming skills, and strong desire to become an ML/NLP expert.