Automatic Generation Of Minimal Test Suite Based On LSGA

With the computer technology rapid improvement, the software has also walked into each domain of human society. However along with the software scale unceasing expansion, the complexity of the software also unceasingly enhances and the effective software development becomes more and more difficult. So the software quality and reliability becomes an important task in the field of software engineering.Software testing is a important method to guarantee the software quality and reliability, but it is a complex process and needs to consume the huge manpower, resource and time. Therefore, it is very important that the enhancement of software testing automaticity can guarantee the software quality and reduce the software development cost. Among them, improving the automaticity of the test case production also is a very important step. In view of the universal question in research work, this paper proposes automatic generation of minimal test suite and improved genetic algorithm to apply in the question.Starting the research work from this question. First, this paper systematically introduces the research of automatic testing, software testing theory and the automatic generation technology of test case. In which, it emphatically elaborates the automatic generation technology of code-based test case, then elaborates the genetic algorithm theory and the realization method.This paper emphatically introduces the realization technology of the system, which is automatic generation of minimal test suite based on LSGA. Combining the test target, which achieves statement coverage, with basic path testing technology and program instrumenter, the author proposes the largest steady genetic algorithm (LSGA) and applied it in the automatic generation of minimal test suite. Then the paper emphatically elaborates the crucial technology of the system realization, fitness function build, path-numbering, parameters coding, selection operator, crossover operator, mutation operator. The author proposes the ?°the neighbor first?±principle and ?°nearby-path numbering?±principle when fitness function is built and path is numbered. Finally the simulation experiment and the performance analysis are carried on the algorithm.The paper applies this theory to develop the prototype system and its performance is obviously superior to simply genetic algorithm. Finally, the paper is summarized and the next research direction is proposed, according to the deficiency which exists in the system.