Möbus, Claus and Seebold, Heiko (2005) A Greedy Knowledge Acquisition Method for the Rapid Prototyping of Bayesian Belief Networks. In: Artificial Intelligence in Education. Artificial Intelligence in Education . IOS Press, Amsterdam, pp. 875-877. ISBN 1-58603-530-4

- Published Version

Volltext (97Kb)
Official URL: http://dblp.uni-trier.de/rec/bibtex/conf/aied/Mobu...


Bayesian belief networks (BBNs) are a standard tool for building intelligent systems in domains with uncertainty for diagnostics, therapy planning and user-modelling. Modelling their qualitative and quantitative parts requires sometimes subjective data acquired from domain experts. This can be very time consuming and stressful - causing a knowledge acquisition bottleneck. The main goal of this paper is the presentation of a new knowledge acquisition procedure for rapid prototyping the qualitative part of BBNs. Experts have to provide only simple judgements about the causal precedence in pairs of variables. From these data a new greedy algorithm for the construction of transitive closures generates a Hasse diagram as a first approximation for the qualitative model. Then experts provide only simple judgements about the surplus informational value of variables for a target variable shielded by a Markov blanket (wall) of variables. This two-step procedure allows for very rapid prototyping. In a case-study we and two expert cardiologists developed a first 39 variables prototype BBN within two days

Item Type: Book Section
Uncontrolled Keywords: structure learning, rapid prototyping, Bayesian belief networks, pairwise causal precedence judgments, greedy algorithm, transitive closure, Hasse diagram, Markov blanket, diagnostic assitance for cardiology
Subjects: Generalities, computers, information > Computer science, internet
Philosophy and psychology > Psychology
Divisions: School of Computing Science, Business Administration, Economics and Law > Department of Computing Science
Date Deposited: 22 Sep 2014 11:04
Last Modified: 22 Sep 2014 11:04
URI: https://oops.uni-oldenburg.de/id/eprint/1916
URN: urn:nbn:de:gbv:715-oops-19974

Actions (login required)

View Item View Item

Document Downloads

More statistics for this item...