Bayesian network learning algorithm paper accepted more here.
November 2006
Continual development of Bayesian network learning algorithm.
March 2006
Completion of PathNet.
November 2005
Completion of DemNet.
This research is funded by the Engineering and Physical Sciences Research Council under grant numer GR/S78148/01.
Basic research overview
About
This research project is concerned with investigation and development of computer decision support systems to assist with health decision making. We predominantly apply our research to dementia diagnosis, although the techniques and methodologies developed are flexible and can be ported to different medical areas and various different domains.
Aims
In part, this research aims to investigate the usefulness of Bayesian belief networks for the purpose of dementia syndrome and pathology diagnosis. We aim to explore the challenges surrounding the construction of Bayesian networks in general for the purpose of medical decision making and for the purpose of dementia diagnosis. Also, we aim to apply the techniques developed in this research to another clinical area: child birth.
Approach
In general, Bayesian networks are constructed either by hand where a domain expert is required to define the domain variables and their relations (structure) and specify conditional probabilities (parameters), or they can be learned from a dataset. Firstly, we shall create, by hand with a domain expert, a Bayesian network models to facilitate medical decision makers in the complex task of dementia syndrome syndrome (DemNet) and pathologies which may coexist (PathNet). Secondly, we shall apply Particle Swarm Optimisation to the problem of Bayesian network structure learning. It will then be possible to compare the models and proecsses developed by hand with those associated with learning from data.