Tabu Search and Genetic Algorithm for Generating Test Data of Class Testing

A Class is not only a template of objects with the same data structure, but also it could be deployed as component and web service in J2EE and .Net platforms. Therefor, Class testing is the key to ensure the quality of a distributed application. Genetic algorithm(GA) has been used to generate test data for class testing. But it has some defects, for example, some individuals would be generated and selected repeatedly, and results in early convergence. Meanwhile, existed genetic coding only deals with the problem with basic data types, and not support user defined data types sufficiently. To solve these problems, this paper proposes the tabu search and genetic algorithm (Tabu GA) for generating test data of Java class, which forbids individuals which have already been executed being performed again. The method also presents a coding scheme, which is one-to-one code between problem and its solutions and could present the data of user defined data type. It extends the adaptable range of generating test data with genetic algorithm. Besides, the termination condition of the algorithm doesn´t depend on the calculation of fitness. Experiments show that the tabu search and genetic algorithm could get better the convergence and the test coverage rate of test data than simple GA.