The species per path approach to GEMGA-based test data generation

This paper discusses the use of Species per Path approach[1] and gene expression messy genetic algorithm (GEMGA) for automatic software test data generation. This research finds another path problem and extends Species per Path approach on dynamic test data generation. In increasing the search space by program transformation for the path potentially suffering from the path problem, this research differs from previous Species per Path. Transforming the program under test can factor out and increase several paths to reach the same target. As a result, each species uses a fitness function tailored for the space for the path. All together the effort of the fitness functions can guide the search to reach the target. The function is minimized by using GEMGA. The work describes the implementation of Species per Path Approach to GEMGA-based approach and examines the effectiveness on a TRITYP program. Compared with other approaches, the experimental results show that it can generate higher quality test data more efficiently, and should be applied to larger applications.