Automatic Coverage Evaluation for a Medical Expert System

Abstract

It is always important to test software products to ensure that they are correct and are built according to the specified requirements. Testing may not only occur after developing the software product. In the other stages of the software development process (e.g. requirement and design stages) testing may occur to detect errors early which can be more effective in saving project scarce resources. In this paper, an expert system is built for dental clinical treatment. A graph is automatically generated based on a formal model built for possible diseases and symptoms related to dental clinics. Later on, a software application is built to automatically generate, execute and verify test cases from the dental expert system formal model. Coverage is also evaluated based on the graph: nodes, edges and paths. Initial test results showed problems with the developed formal model in producing several dead lock paths. Several cycles of improvements to the model were implemented based on the test output and coverage results. Results showed that such approach can be very useful in evaluating expert systems in general.