Pair-wise test coverage using genetic algorithms

There has been an emerging trend to develop software using different components. In this way the cost of the software reduces and the developer is able to complete the system efficiently. The components' code may or may not be visible to the developer. Testing, in this case, requires the development of a set of test configurations that can be applied on the software. However, for software that comprises a large number of components, it is infeasible to test each and every test configuration within the limited testing budget and time. In this paper we propose a GA-based technique that identifies a set of test configurations that are expected to maximize pair-wise coverage, with the constraint that the number of test configurations is predefined. Although the paper primarily focuses on the interaction between software components, the idea can be applied to single code component testing. We performed some experiments using our proposed approach. The results were promising.