Abstract
In testing object-oriented programs (OOP), a fundamental problem of a genetic algorithm (GA) is how to achieve high code coverage for a class under test (CUT), and it is challenging since it requires the search to generate method calls that reach as many different CUT states as possible. This paper introduces a Guided Genetic Algorithm (GGA) that uses static analysis to set the initial population and guide the mutation operator. GGA uses static analysis to generate test cases that focus on only methods of a CUT that help to cover target branches. GGA also uses a seeding strategy for constant values and reference types to increase the likelihood of covering a target branch. We compared GGA with EvoSuite and pure Random Testing (RT) on six Java open-source projects. Our results show that GGA achieved higher branch coverage.