Research For Testing Data Automatic Generation Methods Based On Genetic Algorithms

Software testing is used to find software errors and defects, so as to achieve the goal of guarantee the quality of software. Software testing is an important link in the whole process of software development cycle. The appropriate test data generation is thekey in the process of software testing. Manual constructions of software test data are a tedious and blindly work. It often leads to the low test efficiency, extending the testing cycles; testing adequacy is difficult to secure. Automatic generation of test data can be effective to improve the process, reduce test cost, and improve test efficiency.Automatic generation of test data and testing adequacy is the two basic problems in automated testing. Path coverage is a strictly coverage standard, and if path tests is performed early in the software development, defects and errors can be find early in the software, thus path tests have higher practical value. Completely paths test in the actual software testing is not feasible, and the usual approach is to select a particular path or according to certain standard choose some target paths. Based on these this article has carried on the related research and the main work includes:1. The basic theory and methods of the software test and genetic algorithm (GA) has been analyzed and researched. This article introduced the outset on the relevant theories of software testing and the test data generation method, and then introduces the flowchart of genetic algorithm, the characteristic and its superiority, etc.2. The methods of test data automated generationbased on Genetic algorithm has been researched. In the existing test data generation method and genetic algorithm (GA) theory, this paper introduces the application of genetic algorithm to automatically generate test data model, and focuses on the realization of a parameter coding and genetic operation. Also, according to the specific test data generation problem, the proper selection of fitness function was designed and the genetic operators were improved.3. The improved genetic algorithm was adopted in the automatic test data generation model, such as static analysis, program instrumentation as an auxiliary, according to the selected target paths the test data of covering them is calculated. According to this model, the simulation experiment was carried on and the single path branch coverage and multipath branch coverage automated test datageneration was implemented, and the effectiveness of the model generation path coverage testing data was verified.4. Non-numeric test data-the characters and the class object test data generation base on genetic algorithmhave been researched.The numerical test data generation method has been studied.For the non-numeric test data, the characters and class object test data generation have been studied; And for the problem of the class objecttest data generation, a kind of new class object coding method was design, which can be applied to the genetic operation of class object, and the generalized hamming distancewas applied to the construction of fitness function about the class object.