Search

In this paper we discuss ways in which coverage analysis, obtained during execution of test cases over a rule-base, can be used to highlight problems in both the test suite and the rule-base, thereby pointing to areas in which we cannot guarantee or predict the system's performance. In particular, we present a series of heuristics which use coverage information and meta-knowledge about the larger population to select additional test cases from the population, in the event that the initial test set is incomplete. This forms the basis of an incremental approach to rule-base testing which allows us to both increase completeness of the test set and improve coverage of the rule-base, thereby increasing the kinds of cases for which the rule-base has been executed during testing. We demonstrate this heuristic approach to test data selection using information generated by TRUBAC, a tool which implements the coverage analysis methods, applied to analyze a prototype system for diagnosis of rheumatological diseases. [ABSTRACT FROM AUTHOR]