Genetic algorithm-based test data generator.

摘要

Software testing is meant to increase confidence in the correctness of software. It is a laborious and time-consuming work; and spends almost a half of development resources. Generally, the testing goal is to reveal as many faults as possible, with a limitation on the number of test data to be used. The challenge, in this case, is in being able to minimize the number of test data while maximizing coverage. Obviously, automating the test data generation process is expected to significantly reduce the overall development cost. There are evidences that Genetic Algorithm (GA) has been successfully used in developing test data generators. However, there is no common ground for assessing and comparing these GA based test data generators. In this thesis, based on our critical survey, we present and use a set of attributes for assessing and comparing these generators. Our critical survey has revealed that existing GA-based test data generators suffer from some problems. This thesis presents our attempt to overcome one of these problems; that is the ability to deal with multiple target paths at one time. We have designed a GA based test data generator that is able to overcome this problem. Moreover, we have implemented a set of variations of the generator. Experimental results show that our test data generator is more powerful than others.