Research On Test Data Generation Based On Simulated Annealing Genetic Algorithm

With the rapid development of information technology, a large number of software products infiltrated into every walk of life. How to ensure software quality has become a focus of attention.Software testing is the main means to ensure software product quality and reliability, and its position is no substitute. However, with the development of the software industry, the size of software products are gradually expanding and the complexity of it is increasing. It makes the difficulty of the software testing to be further aggravated. How to effectively reduce the software testing required, which includes the huge manpower, material resources and time-consuming, is the issues that software testing needs to focus. According to statistics, the time-consuming in software testing accounts for about one-third throughout the development cycle, and the cost accounts for nearly fifty percent. Therefore, improving the ability of software test automation is the critical path that ensuring software products quality and reducing software development cost. In order to improve the automation of software testing, how to improve the automatic generation of test case needs to be focused. This paper will focus on research and analysis on this issue.This paper has first briefly introduced the background and significance of studying software testing, and about the situation of studying, and then briefly presented the contents and the structural of this paper.Then, this paper carried out a detailed overview of software testing, and introduced the related technologies about the automatic generation of test case. Meanwhile, their respective advantages and disadvantages are discussed. This paper has mainly analyzed the automatic test case generation for test path technology and some available methods. In this paper, various methods and related techniques have been made a systematic summary and comparison. Then we put forward that the artificial intelligence will play a big role in this area. Subsequently, this paper has described the genetic algorithm and simulated annealing algorithm, and has analyzed the algorithm of their respective principles, elements, the algorithm steps and their respective advantages and disadvantages. As the ordinary genetic algorithm easily falls into the "premature" and has poor diversity of individual in the evolutionary process. This paper has proposed Simulated Annealing Genetic Algorithm (SAGA), and has been used for the automatic test case generation. It can give full play to the characteristics of the two advantages, and improve the performance of the algorithm. This mixed algorithm as the core algorithm of this paper.Soon afterwards, this paper has introduced the system architecture model of test case generation based on SAGA algorithm. Some of key technologies in SAGA algorithm have been improved:such as the coding problem, the branch functions, the program instrumentation way, the fitness function adapting to actual problem, the improvement of the crossover and mutation strategies etc. The improved Simulated Annealing Genetic Algorithm (ISAGA) can not only effectively overcome the "premature" convergence problems, but also greatly improve the efficiency of search.Finally, in the part of experiment, this paper has first implemented for generating test case tool based on the improved Simulated Annealing Genetic Algorithm (ISAGA). This tool uses a program as a classic example and automatically generates test case. Through the analysis of experimental results, it proves that this core algorithm can effectively generate test cases meeting the objectives, and its convergence rate is high, and can overcome the "premature" convergence problems. And then, this paper has made use of several pure mathematical functions with complex mathematical characteristics as fitness function to search for test cases meeting the target of problem. The experimental results show that the proposed ISAGA algorithm has better performance than the general GA algorithm and SAGA algorithm in generating test cases meeting the requirements.