A Method Dependence Relations Guided Genetic Algorithm

Search based test generation approaches have already been shown to be effective for generating test data that achieves high code coverage for object-oriented programs. In this paper, we present a new search-based approach, called GAMDR, that uses a genetic algorithm (GA) to generate test data. GAMDR exploits method dependence relations (MDR) to narrow down the search space and direct mutation operators to the most beneficial regions for achieving high branch coverage. We compared GAMDR’s effectiveness with random testing, EvoSuite, and a simple GA. The tests generated by GAMDR achieved higher branch coverage.