Browse by author
Lookup NU author(s): Eshan Rajabally, Emeritus Professor Pratyush Sen, John Dalton
Full text for this publication is not currently held within this repository. Alternative links are provided below where available.
Bayesian Belief Nets provide a formalism for reasoning under uncertainty based upon probability theory that dates back to the 18th century. They have been used to promote understanding and support decision making in a variety of circumstances. However, the use of Bayesian Belief Nets in many complex problem domains is not a straightforward task and the aim of this paper is to provide a survey of aids to Bayesian Belief Net construction. Firstly, attention is drawn to the range of computer based applications that are available to assist in the application of Bayesian Belief Nets. Following this, direction is offered on the initial task in Bayesian Belief Net construction of determining the topology or graph structure for a particular domain. Guidelines are then given on the important step of quantifying the determined Bayesian Belief Net or populating its nodes with the required probability values. Firstly, techniques to reduce the magnitude of the quantification task are considered. Secondly, various ways of eliciting the required beliefs are surveyed. Finally, means to enhance the rigour, accuracy and objectiveness of elicited beliefs are offered. The presented insights have arisen from pursuing the use of Bayesian Belief Nets for problem solving within BAE SYSTEMS and contribute to the development of software for decision support in systems engineering.
Author(s): Rajabally E, Sen P, Whittle S, Dalton J
Publication type: Conference Proceedings (inc. Abstract)
Publication status: Published
Conference Name: 2nd International IEEE Conference on Intelligent Systems
Year of Conference: 2004
Pages: 457-461
Publisher: IEEE
URL: http://dx.doi.org/10.1109/IS.2004.1344793
DOI: 10.1109/IS.2004.1344793
Library holdings: Search Newcastle University Library for this item
ISBN: 0780382781
