Research On Path-wise Automation Generation Of Test Data Based On Simulated Annealing-Quantum Genetic Algorithm

With the increasing scale and complexity of software products, the research of automatic software testing technique was given more and more attention by academia and industrial community. Automatic generation of test data played an important role in automating tests, which was realized that could radically reduce the cost of software and assure the quality of software. At present, many domestic scholars have done a great deal of research on this subject. Although there are a number of algorithms and techniques which have been able to automatically generate test data, but they are not mature yet, and need to further investigation and improvement.The basic path testing is a very effective method to generate test data through a lot of research and analysis, but the method requires a high performance evolutionary algorithm as the basis. Therefore, this paper proposes to employ the quantum genetic algorithm as the core algorithm, and analyzes the disadvantages of which is used to achieve test data automatic generation, and then improves its shortcomings. Firstly, the coding of quantum chromosomes changes from vector coding to angle coding, which only stores one real number, and greatly reduce the storage space; Secondly, a new dynamic rotation angle strategy ensures that the excellent characters in the population can be better genetic, the offspring find the optimal solution quickly; Thirdly, Hadamard door mutation strategy is adopted to increase the diversity of population; Finally, quantum genetic algorithm and simulated annealing algorithm are combined effectively, which solve the phenomenon of "premature convergence" successfully. The improved algorithm is named "Simulated Annealing-Quantum Genetic Algorithm", referred to as the SA-QGA. Test results based on typical complex continuous functions show that SA-QGA is feasible and effective.At last, this paper designs and achieves a framework of Path-wise test data automatic generation based on the SA-QGA. The framework is divided into three parts:the test environment module, SA-QGA module and test run environment module. The experiment shows that the system can generates test data covered the specified path accurately, which is verified by the triangle classification procedures, and the results show that the system has better searching and convergence performances.