As part of our British Council Newton Fund award, we were extremely pleased to welcome a host of researchers from both the UK and Malaysia to our Big Data and Dengue Fever Workshop, held at the University of Nottingham. In partnership with University Malaya, this 3-day event featured a range of talks, demonstrations and brainstorming sessions about how dengue fever research can be impacted positively by Big Data techniques. Full details of the event below…
Dengue Fever Workshop Programme
10-13th August 2016
UM Welcome talk (Siti Hafizah Ab Hamid, Universiti Malaya)
After a warm welcome from the chair, Dr James Goulding, Dr Siti Hafizah of University Malaya introduced the background to this collaboration, and the programme of work being undertaken by Computer Science in the field.
Background to Dengue Fever (Rafdzah Ahmad Zaki)
We were very grateful to Dr Rafdzah Ahmad for then presenting an extremely informative talk on the nature of dengue fever from a medical perspective, and in particular the climate and environmental features in Malaysia that influence its spread. As well as giving a very thorough review of Dengue incidence and the impact of the disease (including the triangle of spread between the disease, mosquitos and humans), this led ideally into discussions on how improvements in prediction of outbreaks can be made.
Presentation on Modeling Dengue Outbreak via Recurrent Neural Networks (Abdullahi Adamu):
Dr Abdullahi Adamu then followed up on the prior epidemiological discussions, by presenting his first steps in collating both climate and medical incidence data on the disease in Malaysia, and deep learning models for prediction of outbreaks. While this illustrated early work on the Grant, there was clear evidence of the potential of a data-driven approach.
Presentation on Neodemographics (James Goulding):
As part of the Knowledge Transfer component of this British Council Newton Fund grant, Dr Goulding discussed new developments in big data techniques, and in particular novel approaches being developed here at the University of Nottingham. This included discussion of time series prediction models, and in particular the impact that external data sources (such as Call Record Data) can have on epidemiological models.
Presentation on Geospatial Science in Emerging Economies (Mark Iliffe):
A key part of any analysis based around human behaviour is the geospatial components. In this presentation Mark Iliffe led the workshop through prior work undertaken in the group in East Africa. Focussing on Tanzania, Mark addressed the difficulties of employing data collection technologies, and the benefits that drone based sensors can have on analysing enivoronmental factors, especially in urban areas.
Presentation on Understanding Movement via CDR data (Gregor Engelmann)
Gregor Engelmann next presented his work on harnessing human behvioural datasets to analyse mobility patterns. It quickly became evident that such patterns have a high affect on the spread of dengue fever which, while borne by mosquitos, is transferred between areas via human movement. As such, data such as that collected via mobile phone traces is likely to have a particularly significant effect on the accuracy of predictive models for dengue spread.
Presentation on Big Data Analytics (Gavin Smith)
Next Dr Gavin Smith explained how transactional ‘big data’ represents an unparalleled resource for understanding human behavior. However, due to the sheer scale of such data extraction of actionable inferences presents significant challenges. He discussed how (in order to provide actionable intelligence) algorithms are required that process and summarize that data in some form, enabling useful and actionable insights to be drawn. He then presented a novel algorithm for this summarization based on dynamic topic modelling, leading to discussion about how advances in data science might be better harnessed in the medical sciences.
Presentation on Deep Learning from Drone Imagery (Mercedes Torres)
Our final talk of the workshop came from Dr Mercedes Torres Torres, of the Machine Vision lab, who is working with the team on automated land-use recognition from Drone Imagery. This excellent talk on the ability of Convolutional Neural Networks to deliver highly accurate results, raised lots of interesting discussion on the ability of such techniques to identify stagnant water, so often the source of dengue infections.
Paper Brainstorming / Discussions for Future Funding / Next Steps
Rounding off the workshop came some break out sessions on how to move Dengue prediction forwards. A clear consensus emerged in that:
- Continued collaboration was mutually beneficial to both UoN and UM.
- Integration of Predictive Models from machine learning was highly useful to epidemiology.
- For Dengue Fever in particular understanding the movements of the population was key – and as such forming a research relationship with a telecommunications provider would be a key next step in advancing this work.
Many thanks to all who attended this highly productive workshop!