An evolutionary algorithm for selection of test cases.

Applying tests to an implementation to check its correctness is often expensive and may require an excessive amount of time. Hence, it is necessary to find a relatively small subset of tests able to detect as many errors as possible. In this paper, we study several approaches to choose such subsets of tests based on their capacity to detect faults. These faults are defined as mutation operators that are applied to the specification of the systems with the goal of simulating faulty versions called mutants. The different methods are evaluated to determine the ones that provide the best test suite according to the relevance of the tests, their capacity to detect fault and the number of inputs involved. All the algorithms proposed have been implemented in a tool freely available.