Bioinformatics Complex System Identification

Complex modelling problems, like those found in bio-informatics and environmental systems, have the properties that observation data is often sparse and the underlying systems are high-dimensional, non-linear as well as being poorly understood. Modelling such systems involves an iterative refinement process, which is costly and time consuming and often traditional system identification tools may not be suitable for these large, complex systems. This project will develop new iterative model identification techniques based on data mining that are appropriate in such cases. These algorithms can help to develop new interactive model identification techniques based on data mining technologies and hybrid optimization methods suitable for systems with low quality measured data, like those found in typical bioinformatics problems. The efficiency of the model identification phase will be enhanced by a new experimental design framework which governs the data collection and model building process. Decision support tools will be developed to support both off-line design and incremental, real-time applications.