Test data generation method based on multiple convergence direction adaptive PSO

Automated test data generation is a traditional technique for reducing the cost and time of software testing. Various metaheuristic techniques have been successfully applied for this task. In contrast to the typical metaheuristic algorithms applied for branch and path coverage, this study focused on low resource consumption and efficient information coverage for critical path coverage. First, we combined the characteristics of branch coverage and path coverage to determine a critical path based on quantified path scores. As a result, we constructed a fine-grained fitness function based on the uniform scale branch distance. Second, we proposed an adaptive particle swarm optimization (MCD-APSO) algorithm with multiple convergence directions to accelerate convergence and escape from local optima. The proposed MCD-APSO algorithm improved the global search ability by enriching the diversity of the particle swarm and enhancing the current evolutionary information use of the particles. Finally, to validate the performance of the MCD-APSO algorithm, we compared the proposed algorithm with six test-data generation algorithms on six normal-scale and six large-scale benchmark programs. The results showed that the MCD-APSO algorithm outperforms the benchmark programs regarding the mean number of iterations, total running time, and coverage failure probability.