Along with John Paul Gosling, I run UQ Consulting Ltd, a small consultancy that helps businesses develop uncertainty quantification methodologies. This ranges from expert elicitation in substantial practical applications, to more sophisticated mathematical treatments of uncertainty, particularly in the case where there are complex models involved.
Previous projects include:
Network Rail (the UK rail operator) employed us for four months to conduct a substantial review and development of the uncertainty quantification methods used within Network Rail. We focussed on their Whole Life Costs and Asset Life Cycle profiles, and developed methods to account for uncertainties in the models of various asset disciplines, with a focus on their track, signalling and earthworks. Our methodology was used to create the predictions they submitted to government for CP6 (the 5 year long control period starting running 2018-2023).
Many companies have big datasets which they would like to use to make predictions about the future. The past decade has seen an explosion in the ability of statistical models to learn from data. Gaussian processes (my primary area of research) are one of the key machine learning methods to have emerged, and have had great success in making predictions from data (while also quantifying uncertainty in those predictions).
Recent consultancy work has included using Gaussian processes to build an intelligent prediction system for tide levels. By combining historic measurements with the information contained in weather forecasts, it is possible to make more accurate predictions than were previously available.
You can read about my Gaussian process work here.
Many organisations, both public and private sector, are increasingly seeing data as a key corporate asset. Whilst they may have very able and numerate technical staff, some additional statistical training can help companies realise the potential value in their data. Recent work includes providing several days statistical training to an energy engineering consultancy, recapping statistical basics, and showcasing recent developments in machine learning and statistical modelling.
If you have a project you would like to discuss, please get in touch.
Dr Richard Wilkinson
School of Mathematics and Statistics
Hicks Building, Hounsfield Road
University of Sheffield
Sheffield, S3 7RH
Phone: 0114 22 23728