The University of Bath is inviting applications for a funded PhD position under the supervision of Dr Matthew Nunes and Dr Alexander Cox in the Department of Mathematical Sciences and Dr Kjeld Jensen from BT Applied Research.
Applicants should be available to start as soon as possible (and before the end of September 2022 at the latest).
This project is in collaboration with BT Applied Research, who are looking to apply the methodology in their analysis of a wide range of statistical challenges across their business. The work in this project will build on an existing relationship with BT, and regular contact and discussion will play an important role in setting the research direction of the project.
The PhD Research:
In modern statistical modelling applications, there can be large quantities of data which needs to be analysed in real-time, and monitored to ensure the stability and accuracy of the underlying models. In many cases, the underlying data can be complex, and statistical modelling of the time series data may require significant expertise to analyse and careful observation to detect changes in the structural relationships or to provide accurate forecasts of performance. The aim of this project is to develop novel statistical methodologies which can automate much of this complex modelling process and provide easy-to-use diagnostic procedures to enable AI-enhanced data analysis across many areas of the organisation. The project is in collaboration with BT, who will provide expertise and examples of real-world challenges.
A key challenge faced by companies such as BT is that business-critical time-series data is collected, along with potential driving factors such as weather or location. These datasets often display seasonalities and complex nonlinear relationships which may also change over time. Deciding which predictors (features) to use in statistical models plays an integral role in predictive and classification ability. The number of such predictors considered for modelling BT data can be in the hundreds, and often it is natural to choose candidates within groups of predictors and within those specified by transformations of features. Up until recently practitioners within BT have fitted such models by hand, which is time-consuming and costly, and thus automating such approaches is vital. Moreover, the end-users of the models are often not experts, and need models to be easily interpreted. In addition, important features in such models may change over time, and thus decisions should be made dynamically. The primary aim of this project is to develop novel methods for intelligently selecting appropriate statistical models for the data, and automatically detecting significant structural changes in the data. A second strand of this project concerns the development of interpretable approximate models for use by non-experts.
This project will look to combine the use of dynamic time series models and appropriate Bayesian inference methodology, for example sequential Monte Carlo/particle filtering. These inference methods are particularly suitable to dynamic models with complex structures and intractable or computationally infeasible calculations, where traditional likelihood-based techniques have difficulties.
Candidate Requirements:
The successful candidate should have a first class or 2:1 degree in Mathematics, Statistics or another relevant discipline. A Masters qualification in Mathematics or Statistics would be beneficial.
Non-UK applicants will also be required to have met the English language entry requirements of the University of Bath.
Enquiries and Applications:
Informal enquiries are encouraged and should be directed to Dr Matthew Nunes on email address [Email Address Removed].
Formal applications should be made via the University of Bath's online application form for a PhD in Statistics.
See our website for more information about applying for a PhD at Bath.
NOTE: Applications may close earlier than the advertised deadline if a suitable candidate is found; therefore, early application is strongly recommended.
Continue with Facebook
