Study And Practice Of Software Testing For Evaluation Of Rule-Based Expert System

With the explosion of software's scale, the quality of software products has become one of factors that affect the development of computer. In recent years, software testing, the most important technique to assure software's quality, has been an indispensable part in producing software. Expert system is a kind of special software, and applying software testing to it's evaluation shows the extension of software testing to artificial intelligence, and it assure the quality of expert system at the most extend.On the basis of a project of the National 863 Program "Testing and evaluation of the agriculture expert system development platform", this paper makes the deep discussion on applying software testing to evaluation of rule-based expert system, and in turn introduces software testing theory and several typical testing techniques, construction theory and the design on testing scheme for rule-based expert system, the research on inconsistent and redundancy inspection and coverage metrics for rule-base, and implemention of fuzzy evaluation system for expert system quality etc.Moreover this paper also discusses two central problem in testing expert system: the study of testing scheme and rule-base testing techniques. For the former, proposes a phase-based, step-based, application-classified testing scheme; for the latter, puts forward an approach on rule-base inconsistent and redundancy inspection, and motivated by structural testing techniques based on data flow analysis, RBEF(Rule-base Execution Flow Chart) is proposed, which shows the coupling and restriction relationship of rules, furthermore,rule-base coverage metrics criteria and test case design methods are brought out.Finally this paper introduces an software testing frame based on RBEF. It includes following functions: setting up RBEF from rule-base, static analysis, fault inspection and designing testing cases; at the same time, It does further explore for the fuzzy evaluation theory of expert system, and implements a fuzzy evaluation system for expert system quality at last. How to develop out more perfect automatic testing tool is our future research and struggling direction.