Automatic Generation Of Test Data Based On Adaptive Genetic Simulated Annealing Algorithm

The rapid development of information technology makes that software products are used more and more in many social fields. So the quality of software products is becoming a serious problem, and people attach more and more importance to it.Along with the continued expansion of the scale of software, the increase of the software complexity, and the use of Orient-Object methods and tools, the difficulty of the software testing is increasing more rapidly than before. The practical experiences of software developing indicates that the software testing account for about 1/3 of the total developing time and about half of the entire expenses in the entire developing cycle. Unfortunately most software testing works are finished by people, because effective test tools are very expensive now. But in fact, test work has the most probability to be accomplished by computers in the entire developing cycle. The reason is that many operations in test-work are repeated, uncreative and can be finished only with serious patience. The characteristic of computers decides that it is the most suitable role to replace people to finish this kind of work.In this paper we concluded and summarized the methods of software testing, and detailedly introduced dynamic testing, static testing, black-box testing and white-box testing. Finally summed up the test data generation technologies in recent years, and according to the systematical sum-up of a variety of methods and techniques, extracted the methods and techniques used in this paper.Next, some main searching algorithms for obtaining test data are researched, and we finally choose Genetic Algorithm the base algorithm in this paper. At the beginning, we analyze both the advantages and disadvantages of Genetic Algorithm carefully. Then, we improve the algorithm by following ways. First, we put the anneal mechanism of the Simulated Anneal Algorithm into the genetic algorithm and improve it. Second, we design the Adaptive Mutation Probability to make sure that the best individual can not be destroyed and reduce"undulation". Third, by using the Fitness Zoom method, we not only solved the easy"premature"problem of Genetic Algorithm, but also accelerated the...