Date of Award


Document Type


Degree Name

Doctor of Philosophy (PhD)


Information Systems and Decision Sciences (Business Administration)

First Advisor

Ye-Sho Chen


Rule-based representation techniques have become popular for storing and manipulation of domain knowledge in expert systems. It is important that systems using such a representation are verified for accuracy before implementation. In recent years, graphical techniques have been found to provide a good framework for the detection of errors that may appear in a rule base. In this dissertation, we develop a technique that uses a directed hypergraph to accurately detect all the different types of errors that appear in a rule base. This technique overcomes limitations of existing graphical techniques that are unable to accurately detect all the errors that appear in a rule base, without misdiagnosing error-free instances. The directed hypergraph technique allows rules to be represented in a manner that clearly identifies complex dependencies across compound clauses in the rule base. Since connectivity across compound clauses are accurately represented, the verification procedure can detect errors in an accurate fashion. We have developed a verification procedure that uses the adjacency matrix of the directed hypergraph. The procedure detects different types of errors by using simple operations on the adjacency matrix. In practice, expert systems are often used to make inferences based on multiple observed facts. Most existing techniques have ignored this aspect, since the selection of valid combinations of rule antecedents from a large number of rule antecedents to be considered is difficult. To address this issue, the directed hypergraph technique has been extended to perform verification checks when sets of feasible multiple assertions are made available to the system. As the size of the rule base increases, execution of the algorithm can be hard due to storage and computational considerations. It has been empirically found that sets of rules in large rule bases are sufficiently separated to allow decomposition into smaller sets. The directed hypergraph technique has been enhanced to accurately detect all errors in large rule bases by performing verification checks over the smaller groups of rules separately, and propagating the results from one group to other linked groups.