(An) empirical evaluation of test data generation techniques

Abstract

Software testing for assuring the software quality is known to account for approximately 50 percent of the development cost. This cost can be reduced if the process of testing is automated. But, automatic selection and generation of test inputs still remains a challenge for tool developers. Therefore, a lot of works have been performed to automate the process of generating test data. Although many techniques for automatic test data generation have been developed, [9, 14, 17, 23, 28, 31, 36, 38, 48, 49], there is no overall evaluation and comparison of these techniques. Evaluation and comparison of existing techniques is useful for choosing appropriate approach for particular applications. Evaluation and comparison of existing approaches also provides insights into the strengths and weaknesses of current methods, and offers a guidance in choosing areas that future work on the test data generation should address. This paper discusses on the issues relevant to test data generation and conducts experiments on four representative test data generation techniques. The results of the experiments show that the GA-based test data generation performs the best. However, there are still some weaknesses in the GA-based method. Therefore, we introduce the static analysis information into the GA-based method to cope with these weaknesses. The experiments are carried out to compare the original GA-based method and new version of the GA-based method that utilizes static analysis information.