| Software testing is essential to ensure software quality. Some kinds of software which have many key outputs deserved attention are more suitable to develop test cases from output domain. If test automation could be put in practice, the cost of software testing would be significantly reduced. The key element in test automation is the automatic generation of test inputs and expected outputs. However, for some given inputs, automatically generating expected outputs according to specifications is difficult. Therefore, it is important to explore a method for output domain based automatic test cases generationMost kinds of software have nonlinear relationship between inputs and outputs, and the neural network can map nonlinear relationship admirably. Thus, this paper uses neural network to create a function model that can be taken as a function substitute for the software under test. Then based on the created function model, for given outputs, genetic algorithm is employed to find the corresponding inputs. The new crossover operator and mutation operator are presented to improve the genetic algorithm. The experimental results indicate that the approach of output domain based automatic test cases generation is promising and effective, the created function model can simulate the function of the software under test perfectly, and the improved genetic algorithm can enhance the efficiency and successful ratio of test case generation at a large extent. |
