Evolutionary Generation of Test Data for Multiple Paths Coverage

Generation of test data using genetic algorithms has attracted many researchers' interests in recent years, the efficiency of previous methods, however, needs to be further improved. A method of generating test data based on genetic algorithms to cover multiple target paths in one run is presented in this study. First, the problem of generating test data is formulated as a multi-objective optimization problem in which the number of objectives decreases along with generation of test data. Then, test data are generated using genetic algorithms incorporating with domain knowledge. Finally, our method is applied to two typical benchmark programs, and compared with Ahmed's method and the single path method. The experimental results confirm that our method has advantage in terms of the number of generations and the time consumption.