Weak Mutation Test Set Generation Based On Dynamic Set Evolutionary Algorithm

Mutation-based Test Case Generation (MTCG), which has become more and more important in software testing, has attracted much attention from academia. The mutation test is a kind of software testing technology based on fault instrumentation. The mutation operation is used to change the grammar of the original program to generate mutant. MTCG technology is designed to generate very few test cases to kill as many mutants as possible to achieve a higher quality test case set. However, the mutation test is too expensive to be applied to the actual test of the software. Therefore, how to reduce the mutation execution cost is the key research content of MTCG technology.Weak mutation test set generation based on set evolution is an effective method that different test sets are made as individuals and all mutant branches are made as objective to generate test set, but there are two problems: First,the size of the individual (the number of test cases included) in the evolution of the process is not changed, however, the the suitable scale of individuals is difficult to set. Second, the used fitness function has high computational cost,and is not suitable for the set individual in the evolutionary method.Therefore, a new method called Dynamic Set Evolutionary Algorithm was proposed. In the evolutionary process, a reduction operator on those sets is presented to adjust the scale of individuals according to execution information, avoid the influence of individual initial scale. By the way, a suitable fitness function has been designed to evaluate those test sets, and the merits of the individual are evaluated more accurately. At the same time, the computational cost is less and the efficiency of the algorithm is high. In addition, the population update takes the (?, 1 +?) update strategy by keeping the optimal individual in parents to guarantee population update quality. The above method makes the weak mutation test set smaller and generation time lower under the premise of satisfying the weak mutation test criterion.The results of this paper show that compared with weak mutation test set generation based on set evolution method, this method cannot be affected by the individual initial size and generate smaller weak mutation test set in less time under the premise of satisfying the weak mutation test criterion. The minimized size is about 50.15% on average than the initial size of the individual. And the execution time is also lower than the original algorithm,which has a maximum reduction of 74.58%. Therefore, this method provides an executable solution for weak mutation test set generation.