The University of Hull is offering one UK/EU PhD Scholarship for candidates applying for the above
project.
Closing date: Friday 27 May 2011.
Studentships will start in September 2011
Supervisor: Dr Chandra Kambhampati
Email: c.kambhampati@hull.ac.uk
Telephone: +44 (1482) 465744
Department: Computer Science
Co-Supervisors: Professor John Cleland (Cardiology), Dr Darryl Davis (Computer Science)
PhD Project: http://www2.hull.ac.uk/science/computer_science/prospective_students/postgraduate_research/projects.aspx
Full advert at:  http://www2.hull.ac.uk/science/pdf/CS%20Chandra%20Ad%20for%202011%20PhD%20Studentships.pdf

Clinical data often presents itself with a number of problems, e.g. high dimensionality, missing values and unbalanced classes of data amongst others. These create a unique set of problems for developing appropriate classification/prediction models, and compromise the development of Decision Support Systems. Often, the first stage is to reduce the dimensionality of the problem. Dimensionality reduction techniques can be categorised into two (a) feature extraction and (b) feature selection techniques. The choice of technique is dependent on the nature of the final solution that is required. For example, in image processing feature extraction could be a very attractive option. On the other hand developing decision support systems, which will use the variables observed, in order to make sense to the user, feature selection techniques could prove to be more amenable. With clinical datasets, the resolution is not straightforward. The efficacy of the computational strategy is dependent on the manner in which missing data is handled. The methods used for this, dictates the methods used for handling unbalanced classes. The project proposes to investigate and develop methodologies which would be able to handle high dimensionality, missing values and unbalanced classes.

The aims of the research are thus
(a) Investigate methods for handling missing data e.g. Expected maximum likelihood, imputation, 
hot-decking, mean/mode, classifier and ensemble methods
(b) Investigate the relationships between methods for filling in missing data feature 
extraction/feature selection methods, with a view to develop prediction models and Decision
Support Systems
(c) Develop Bayesian methods for missing data and thus develop an integrated solution for handling
missing data and feature selection, and at the same time handle unbalanced classes.

This full-time UK/EU PhD Scholarship will include fees at the ‘home/EU' student rate and maintenance
(£13,490 in 2011/12, subject to final confirmation) for three years, depending on satisfactory 
progress. PhD students at the University of Hull follow modules for research and transferable 
skills development and gain a Masters level Certificate, or Diploma, in Research Training, in 
addition to their research degree.

If you require to know more, please contact the Principal Supervisor: Dr Chandra Kambhampati via
Email: c.kambhampati@hull.ac.uk or myself (as Co-supervisor or as Director of Postgraduate Research Studies )

Darryl N. Davis
Director of Research
Director of Postgraduate Research Studies