Research On Automatic Generation Of Test Data Based On Genetic Algorithm

Software testing is the key of guaranteeing software quality. It is a process which is indispensable in software development. It also is an important part of software engineering. The effect of software testing is relating the quality of software production directly. But software testing is a very complicated process, it needs huge of people, matter, money and time, the test engineer must design, excute and analys large numbers of testing examples. Most of these are carried out by handwork. So the efficiency is low. And that also makes the cost of testing on a high level.Automatization of software testing is a work which do our preconcerted plan using automation test tools, script or other methods. The technique of automatic generation of test data is one of the most important technique. We can generate test examples using this technique for the program . Realization of this technique will change the traditional method which generate test data by intuitions and experiences. That will improve the efficiency of software testing, and lighten the work we do when we are writting large number of test data. So, researching the technique of automatic generation of test data and carry them out has a very important meaning for automatization of software testing.Genetic algorithm is a optimization method which simulates heredity and evolution of biology. It put behaviors similary with genetic genes such as crossover, recombination, mutation, selection and so on into the progress of algorithm. One of the characteristic of genetic algorithm is it not only keep the best individual from the part, but also get better individuals by using crossover, recombination or mutation. Since the nineties of last century, people start trying genetic algorithm and other artificial intelligence techniques into sofeware testing, and obtain certain of research fruit.This article basis the request of automatic generation of test data, combine the theory and the characteristic of genetic algorithm, modify parts of the genetic algorithm then apply into software testing. It confirm the coding method for most familiar types of test data: numerical value type, non numerical value and class object type. It also select fitness functions, improve the method of data output, actualize the automatization of software testing. At last, it contrast some of the methods used in this article with other methods by experiment analyse. It especially analyse the number of the test data they put out and the veracity of the data .And these analyses prove that this article has certain advantage in these aspects.