Genetic Algorithm In Path Coverage, Research And Application Of Test Data Generation

With the development of software technology and the increasement of software project scale, the effect of software testing becomes more and more important.In testing, the selection of test data is a nodus to structure testing. Whether the errors of programs can be detected or not are directly related to whether the test data is right or not. Although some methods are brought out to automatically generate test data, in the practical application there are no perfect solutions because of their localizations. The test data can be gained only by the experiences. This paper focuses on the method of automatically generate test data using Genetic Algorithms (GA) in path coverage testing.At the beginning, this paper introduces the software testing technology and Genetic Algorithms (GA).On the software testing technology, we introduce the concepts, classes and phases, investigates the methods of selecting proper test data, and emphasizes on the various of the methods which use path coverage testing. On the GA, we introduce the generation development basic concept feature and basic glossary, expatiates the general process of GA, analyse the important factors which effect GA.Secondly, the paper proposes the application method of GA used in test data generation. We analyse the academic gist of test data generation using GA and the possibility of using GA. Then we discuss some problems needed to be considered using GA generate test data in path coverage testing, especially on the problem of analysis and expressions of paths, the design method of fitness function, and the method of program instrument ion, we improve on them according to the practical requirements. At last we use a simple example to show the process of test data of basic type generation with GA.The paper deeply investigates a difficult problem in test data generation with GA - the generation of class-object test data. First we introduce a new method of coding class-object to make it suitable for genetic operations. Then we analyse the Extended Harmming Distance (EHD), and proposes the fitness function for class-object generation based on EHD. At last we use a simple example to show the process of class-object test data generation with GA.Lastly, the paper uses the method in the Personal Loan Subsystem of Bank of Communiction in the Unit Test phase. We create a tool model to automatically generate test data according to the requirement and give out the experiment result. We analyse the results and compare them. The results prove that the method of generating test data in path coverage testing using GA is agility, efficient. The method has some practical values and is valuable to be investigated further.