Models Of Test Data Generation For Two Types Of Specific Path Coverage And Evolutionary Solutions

Software testing plays an important role in ensuring the quality of software. As the path coverage is a promising research direction, people have already done deep study on it in recent years. For common software, there always contains many paths while these methods consider few paths coverage problem and they cannot resolve these problems effectively. Therefore, an efficient test data generator for multiple and many paths needs to be proposed to meet the requirement of software testing. Besides,for software with random numbers, traditional methods of generating test data often lose their effectiveness. This thesis gives the models of test data generation for two types of specific paths coverage and evolutionary solutions. The main research contents contain the following two aspects:(1) For path coverage problem of common software, this thesis proposes a test data generation method for multiple paths coverage based on local evolution. The optimization model is established, in which coarse-grained functions are used as the objective functions. We propose a GA with local evolution to solve the optimization problem. In evolutionary phase of the algorithm, individuals are divided into several groups according to the objective functions. Then evolutionary operators are implemented to the individuals in each group. After evolution, the individuals are selected according to the fine-grained fitness functions. The efficiency of test data generation is improved by using functions of different granularity in different phases.(2) For software with random numbers, this thesis proposes the adequacy criteria of testing. Based on it, the problem of generating test data can be transformed to an optimization one. We establish the mathematical model of generating test data for software with random numbers and propose corresponding evolutionary optimization solutions. During evolution, it is hard to calculate the individual fitness. This thesis gives the approximate calculation method of individual fitness using the knowledge of probability theory.Results of this study not only enrich the theoretical knowledge of software testing, but also improve the efficiency of test data generation. This study provids technical support for the quality of software.