An Improved Genetic Algorithm for Generation of Pairwise Combination Test Cases

To solve the problem of redundant test cases in the process of generating test cases for numerical simulation software, we proposed Population Split Genetic Algorithm (PSGA). We adopt the idea of greedy algorithm and population splitting as well as individual exchange to improve generation algorithm. Firstly, We introduce the idea of greedy algorithm to update the fitness in genetic algorithm. Secondly, we add the steps of population splitting and individual exchange between populations on the basis of genetic algorithm. Improved genetic algorithm enhances the global optimization ability and avoids falling into the local optimum dilemma when generating test cases. Finally, we proposed an evaluation method based on the redundancy of covered combination. We compared the test case generation results with PICT, Allpairs and Acts tools. Furthermore, we compared with genetic algorithm and its derivatives. Experimental results show that the PSGA can effectively reduce the number of test cases compared with the above tools and algorithms.