Automatic Test Data Generation For Path Testing Using Genetic Algorithms

With the development of informationzation, software products being used more and more widely, the system getting more and larger and complexity, the quality of software becomes a hotspot and difficulty problem of computer technical field. Being an important measure to assure quality and reliability of software, software testing becomes one of the most important aspects of software industry either in our country or abroad.This paper focuses on automatic test data generation for path testing using genetic algorithms. That is a program P and a target path W, D is the set of all possible inputs, the goal is tofind input values x (?) D , that will traverse the target path W .Path testing is a problem of NP. Itis also an important method to test the structure of programs. Studies of automatic test data generation can reduce the high cost of manual software testing and at the same time increase its reliability. So studies of automatic test data generation is a problem of practical meaning for the realization of automatic software testing.First, this paper introduces several steps of test data generator system. They are program analyzer, path selection and test data generation. Main contents of every step and which one in use is also given out in this dissertation. The genetic algorithm is used as the core algorithm of automatic test data generation based on analysis and comparison of the methods such as random algorithm, symbol executing algorithm target oriented algorithm, path oriented algorithm and genetic algorithm.Catch tightly, this paper introduces the basic principle of genetic algorithm simulated annealing genetic algorithm and immune algorithm respectively, and analyzes the advantages and shortcomings of each one. Because genetic algorithm runs short of variety and has the problem of precocity, this paper proposes a new algorithm— Annealing Immune Genetic Algorithm (AIGA) as the core of the automatic test data generation. This algorithm uses the expectation of reproduction instead of fitness function, and use annealing temperature to adjust the expectation of reproduction.At last, as an example, we generate testing data for the Program of Triangle Classifier. The result shows that AIGA can minify the lack of variety by restrain the antibody with high density, then make the crossover operate have a better effect .So with the experiment, the validity of AIGA can be clearly seen.