Software Test Data Generation using Ant Colony Optimization

Software testing remains the primary technique used to gain consumers' confidence in the software. The process of testing any software system is an enormous task which is time- consuming and costly (1). The development of techniques that will also support the automation of software testing will result in significant cost savings. The application of artificial intelligence (AI) techniques in Software Engineering (SE) is an emerging area of research that brings about the cross- fertilization of ideas across two domains. A number of published works, for examples (2) and (12), have begun to examine the effective use of AI for SE related activities which are inherently knowledge intensive and human-centred. It has been identified that one of the SE areas with a more prolific use of AI techniques is software testing. The focus of these techniques involves the applications of genetic algorithms (GAs), for examples (8) and (10). Other AI techniques used for test data generation included the AI planner approach (7) and simulated annealing (13). Recently, Ant Colony Optimization (ACO) is starting to be applied in software testing (3, 10). Namely Boerner and Gutjahr (3) described an approach involving ACO and a Markov Software Usage model for deriving a set of test paths for a software system, and McMinn and Holcombe (10) reported on the application of ACO as a supplementary optimization stage for finding sequences of transitional statements in generating test data for evolutionary testing. However, the results obtained so far are preliminary, and none of the reported results directly addresses specification-based software testing.