Return to search

Implementing Dempster-Shafer theory for inexact reasoning in expert systems

The work described in this thesis stems from the idea that expert systems should be able to accurately and appropriately handle uncertain information. The traditional approaches to dealing with uncertainty are discussed and are shown to contain many inadequacies.
The Dempster-Shafer, or D-S, theory of evidence is proposed as an appealing theoretical basis for representing uncertain knowledge and for performing inexact reasoning in expert systems. The D-S theory is reviewed in some detail; including its approaches to representing concepts, to representing belief, to combining belief and to performing inference.
The D-S implementation approaches pursued by other researchers are described and critiqued. Attempts made early in the thesis research which failed to achieve the important goal of consistency
with the D-S theory are also reviewed.
Two approaches to implementing D-S theory in a completely consistent manner are discussed in detail. It is shown that the second of these systems, a frame network approach, has led to the development of a fully functional prototype expert system shell called FRO. In this system, concepts are represented using D-S frames of discernment, belief is represented using D-S belief functions, and inference is performed using stored relationships between frames of discernment (forming the frame network) and D-S belief combination rules. System control is accomplished using a discrete rule-based control component and uncertain input and output are performed through an interactive belief interface system called IBIS. Each of these features is reviewed.
Finally, a simple but detailed example of an application of a frame network expert system is provided. The FRO system user's documentation is provided in the appendix. / Applied Science, Faculty of / Civil Engineering, Department of / Graduate

Identiferoai:union.ndltd.org:UBC/oai:circle.library.ubc.ca:2429/28383
Date January 1988
CreatorsFroese, Thomas Michael
PublisherUniversity of British Columbia
Source SetsUniversity of British Columbia
LanguageEnglish
Detected LanguageEnglish
TypeText, Thesis/Dissertation
RightsFor non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

Page generated in 0.0016 seconds