Comparison of Two Fitness Functions for GA-Based Path-Oriented Test Data Generation

Automatic path-oriented test data generation is not only a crucial problem but also a hot issue in the research area of software testing today. As a robust metaheuritstic search method in complex spaces, genetic algorithm (GA) has been used to path-oriented test data generation since 1992 and outperforms other approaches. A fitness function based on branch distance (BDBFF) and another based on normalized extended Hamming distance (SIMILARITY) are both applied in GA-based path-oriented test data generation. To compare performance of these two fitness functions, a triangle classification program was chosen as the example. Experimental results show that BDBFF-based approach can generate path-oriented test data more effectively and efficiently than SIMILARITY- based approach does.