Breeding Software Test Data with Genetic-Particle Swarm Mixed Algorithm

Abstract

Software test is usually costly and vital in software development lifecycle. Though genetic algorithms have the globally searching capability, premature convergence and weak local optimization are two key problems existing in the conventional genetic algorithm. This paper introduces particle swarm optimization into genetic algorithm to breed software test data automatically. The GPSMA (Genetic-Particle Swarm Mixed Algorithm) uses the individual’s update mode to replace the mutation operation in genetic algorithm on the basis of population division. The experimental results show the new method can not only maintain effectively the polymorphism in the colony and avoid premature, but also greatly improve the convergent speed.