Evolutionary Generation Of Test Data For Mutation Testing Based On Semantic Analysis

Mutation testing,the technique for evaluating the fault detection capability of the testing date,is being widely adopted for task of software testing.However,there are still many difficult problems that need further study,such as the large number of mutants,the low efficiency of mutation test data generation,and the equivalent mutation.As two of the hot topics regarding mutation-testing task,reducing the amount of mutation and generating testing datasets for removing mutation are researched thoroughly in the dissertation.The overall contribution of the research could be summarized as follows:(1)Semantic analysis(SA)method is proposed for improve the performance of traditional mutation testing problems from the perspective of reducing mutation operators,and the reduction procedure is realized by selecting representative mutation operators.For this reason,the SA based mutation testing model is realized by solving the two objectives,which take both the mutation testing degree and amount of operators into consideration,with genetic algorithm.The SA model firstly conduct the intra-group reductive operation with the overall five kinds of operators followed by reduction on groups.Based on the reduction results,two-stage optimization model is introduced for automatic search of representative operators.The proposed SA model also has the advantage of fast convergence by sequential optimization of representative operators generated from the semantic analysis results.Both the efficiency and effectiveness are considered when constructing the objective function for problem optimization,and the amount of operators could be ensured to be less while keeping the mutation distribution as high as possible.(2)Rebirth-genetic algorithm is proposed for efficient mutation testing data generation.The mutant branches,which are then mixed into the original programs,are firstly constructed on the basis of weak mutation testing theory,and then the corresponding to-be-dealt program are formed.Based on the generation of mutant branches,the specific ranking condition is solved by simplifying the complex operators,followed by obtaining the mathematical models towards testing data generation problems.The models are finally optimized and solved with rebirth-genetic algorithm,and the effectiveness of the algorithm is verified by comparing the solution generated by traditional genetic algorithm.The model proposed in the thesis could not only enrich the mutation testing research,but also improve the efficiency of mutation testing.What's more,we also argue that the established model is of great significance for both theoretical and experimental analysis.