One strand of my research has been on automated fuzzy-rough feature selection and fuzzy rule induction for high dimensional datasets. This work forms a unique approach for dealing with uncertain, vague and/or incomplete data and knowledge. It has been developed with the aim to significantly scale up the utility and user-acceptability of the existing methods. The resulting techniques are applicable to many problem-solving tasks, ranging from learning-from-data and pattern classification, through information retrieval and e-document processing, to systems monitoring and control.
I have also been busy building approximate reasoning techniques for some considerable time. For example, my work on fuzzy qualitative simulation, known as FuSim in the literature, was the first that extended qualitative reasoning from pure symbolic to semi-symbolic. This bridges the gap between qualitative and quantitative simulation. More recently, my team's work on fuzzy interpolative reasoning has led to a novel approach for performing inference in sparse fuzzy rule-based systems and also for the reduction of fuzzy model complexity. A publication which recorded the initial results of this research was awarded the IEEE Computational Intelligence Society Outstanding Transactions Paper Award, in 2009.
One particular focus of my current research is the continuing work on flexible compositional modelling. This is intended to improve the viability of model-based decision-making. A key theoretical foundation for this is preference and prioritisation-based dynamic constraint satisfaction. Based on previous work in this area, an ex-PhD student of mine was awarded a British Computer Society Distinguished PhD Dissertation Award. Supported by the EPSRC we are applying this work to domains such as serious crime scenario synthesis, intelligence data analysis, and terrorism risk estimation. This research offers the potential to improve the use of police and security service resources, thereby reducing the potential for miscarriages of justice.
Much of my research involves collaboration with industrial partners and other end-users, including for instance, British Aerospace, British Geological Survey, Intelligent Applications, Lothian and Borders Police, Marconi Simulations, NCR, Siemens, and Scottish Drug Enforcement Agency. This helps to test and validate both theoretical and applied work. Consequently, this will ensure that the approaches taken are relevant and applicable to real-world needs.