PhD Studentship: Random forests for noisy applications in finance

Country: United Kingdom;

City: Southampton

Vacancy added: 16.05.2019

Employer: University of Southampton

Resumes due: 15.07.2019

 

Project description:

In the financial and insurance sector, machine learning (ML) challenges usually arise either in a setting where data is contaminated by noise or the model to be learned is of a stochastic nature. However, traditional ML models like neural networks are not designed to properly deal with this stochastic component in the data. For this reason, DEVnet (https://www.devnet.de/ ), a leading provider of consulting services for global insurance, energy and financial services companies, is interested in investigating random forests for specific application cases. Examples have shown that traditional off-the-shelf random forest software does not deliver satisfactory results. Therefore, it is intended to extend this promising method to achieve an improved fit and increased reliability. 

In this project we will investigate selected variants of regression trees.  In particular, we will consider one of the major practical obstacles in the day-to-day use of random forests: after calibration it is usually not clear which importance each input feature has. Based on experience from importance concepts used in standard linear regression, we will consider game-theoretic approaches based on Shapley-value like concepts in particular.

All these issues will be investigated both from a theoretical as well as from a practical perspective based on selected problem instances. The selection of the problem instances will come from the project sponsor DEVnet  and cover the application areas finance, insurance, energy and logistics.

DIAMOND: from Data and Intelligence via ModelliNg to Decisions

This project is part of the Southampton DIAMOND initiative of industrially funded PhD projects in Operational Research, Data Science, and mathematical modeling. This year, eight funded studentships are available within DIAMOND.

CORMSIS, the Centre for Operational Research, Management Science, and Information Systems

You will be part of the vibrant research environment of CORMSIS, the Centre for Operational Research, Management Science, and Information Systems at the University of Southampton. CORMSIS has an established breadth and depth in Operational Research unrivalled in the UK. Our research centre applies advanced mathematical and analytical modelling to help people and organisations make better decisions. CORMSIS is the largest Operational Research group in the UK, spanning Mathematical Sciences and Southampton Business School. Among the many areas of expertise, it has extensive breadth and depth of experience in mathematical modelling and optimisation, but covers the whole spectrum of current OR/MS/IS from mathematical optimisation through business analytics and simulation to qualitative research in problem structuring.. In the QS World Rankings by Subject 2019, Operational Research and Statistics at the University of Southampton are placed at 48th in the world and 7th in the UK.

(http://www.southampton.ac.uk/cormsis/)

How to apply:

Scholarships will be awarded on a competitive basis. Applicants should have or expect to obtain the equivalent of a UK first class or upper second class honours degree (and preferably a master’s degree) in mathematics, computer science, engineering or other relevant discipline. The studentship provides a maintenance grant at the Research Council UK rate and tuition fees at the UK/EU rate. Applications should include a cover letter, CV, detailed academic transcripts and the contact details for at least two academic referees.

 
Where to send resume: