Research On Automatic Test Cases Generation Based On Output Domain

With the sustainable development of IT industry, a lot of software products have been widely used in all industries, in order to improve the quality of a software product, the software testing is spending more much manpower and material resources. Software testing is the focus of the design test cases, if test cases could be generated automatically, the cost of software testing would be dramatically cut down. For some kinds of large software, the outputs of them need to be checked, so test cases should be designed based on those outputs. However, if giving the output of software, it is difficult to generate corresponding input through the specification of software. Therefore, in order to improve the efficiency of test cases generation and cut down the cost of software testing, the method of generating test cases automatically from output domain of software tested should be under consideration.This Paper gives the design and research in detail focusing how to use the neural network and genetic algorithm to realize output domain based automatic test cases generation. First, using BP neural networks to establish function model of software to simulate the function of software which can express the relationship between input and output. Then, adopting improved multiple genetic algorithms to search the corresponding input of giving output of software beforehand. So as to realize the output domain based automatic test cases generation, and then in detail describes the coding method, the fitness function, genetic operator and transfer operation of genetic algorithm.In the genetic algorithm improved, this paper proposes an improved genetic algorithm of multiple populations, This algorithm has there populations with different evolution pattern, and the first as a probe and focus on the overall search, the second as a development and focus on the overall search, while the third can balance the search ability at both the local search and local search through the migration pattern, three populations exchange the optimal individuals through the migration pattern and population initialization operation, thus improve the searching efficiency of genetic algorithm, and then the real number coding is used in genetic algorithms, so the individual has a high precision, and the algorithm does not need the process of encoding and decoding, and improve the working efficiency.Subsequently, this paper checks the method which uses the neural network and genetic algorithm to realize output domain based automatic test cases generation. Tree procedures for experiments, the result showed that:the function model could be established using neural networks, and the improved multi-population genetic algorithm can realize output domain based automatic test cases generation, and the improved multi-population genetic algorithm compared with the general genetic algorithm and the double-population genetic algorithm, the results show that the improved multi-population genetic algorithm has improved the efficiency and success rate of generating test cases to a great degree.