Using Genetic Algorithm for Whole Test suite Generation of Object Oriented Programs?

In Software testing there is a great demand for the automation of test cases. The test cases are either generated before coding using software specifications or after coding using program execution traces. Many genetic algorithms have been proposed for procedural programming but they do not suit well for object oriented programming. This is because in object oriented programs the object relationship exists and considering them is important for the generation of best test suite. We propose a method to generate test suites for object oriented programs using genetic algorithm by considering the object-relationships and their dependencies such as polymorphism, message passing and inheritance that exists among them. The key feature of our proposed technique is that the test suites are evolved as a whole instead of generating one test case for each coverage goal. To investigate the effectiveness of our approach we have used a java program of Chocolate Vending machine as the source code and few test cases have been relatively developed to evaluate the fittest test suite with maximum coverage. We have applied our technique in the EVOSUITE tool to determine the efficiency of our approach. Our results indicate that the use of genetic algorithm in Test case Generation is beneficial than the traditional approach of targeting single branches.