There are massive amounts of data presented to the internet in real or near-to-real-time that allow monitoring of economic and societal conditions, amongst other things. This has been used to some effect in the automated monitoring of stock fluctuations, used to inform algorithmic trading. The evaluation of data-sources and curation of these on the basis of predictive power is an area requiring exploration. The project here will focus on a multitude of data sources that would allow the real-time evaluation of geo-political conditions around the globe, with the intention to predict various market shifts and impending political flash-points. Data will be captured by a wide range of sources, including multiple languages, print, audio and video. Text-mining methods will be developed to generate topic models and monitoring these over time, including those emergent. Methods will be explored that evaluate predictive performance for the purposes of curating data-sources and selection of modelling techniques.
The project deals with large data issues, data-mining/machine-learning methods, cloud-computing and interaction with APIs. The project will be heavily computational in either R or python, with potentially compiled languages for computational bottlenecks. The ideal candidate would have a good grasp of practical computing and statistics.
For more information, please see the School's Postgraduate Research page, and in particular the information about Statistics PhD opportunities.