ACO Based Multi-objective Test Case Prioritization And Parameters Optimization

After the modifications of software, regression testing will be conducted to confirm that the functions of the system are not adversely affected. Regression testing is an important part of software testing to guarantee the quality of software. The most common way is re-executing all the test cases which have been performed before to test the software. Because of frequent modifications of the software, constant updating of version and regular correcting the errors, the frequency of performing regression testing also becomes higher and higher. However the cost of executing all the test cases grows sharply with the increase scale of software, and also the number of test cases is rising during the process of software development. Many researchers proposed some optimization techniques to optimize the test suite, including test case selection, test suite minimization and test case prioritization.This paper studies test cases prioritization method for testing optimization. Test cases prioritization is a technology that sorts the test case according to a certain criterion. It was used to optimize only one object in the last few years. But with the changing in the development and improvement of the testing requirement, more than one object should be taken into account in the test case prioritization process. This paper proposes an ACO algorithm for a multi-objective test case prioritization in regression testing which reorders test suite based on maximizing the average percentage of statement coverage (APSC) and minimizing effective execution time (EET). The probability and heuristic function is designed according to the specific problem and the main parts of the algorithm are introduced in detail, including solution construction and evaluation, multi-objective non-dominated sorting, updating set of solutions and updating pheromone.The parameter settings of the ACO algorithm are critical for fast convergence to optimal solutions and are highly problem specific. As multiple parameters optimization itself is a multi-objective optimization problem. A simple’rules of thumb’approach is applied to optimize the parameter settings of ACO to ensure fast convergence to high-quality solutions with a low computational cost. The obtained optimal parameter settings are verified by using a search based parameter optimization approach. Empirical studies are implemented on five subjects from Software-artifact Infrastructure Repository (SIR) and an open source program V8published by Google. The results suggest the ACO algorithm can be applied to multi-objective test case prioritization, and the proper parameters setting, obtained by a simple’rules of thumb’, can offer the trade-off between solution goodness and convergence time, the result of genetic algorithm is not inferior to the former which shows the feasibility and effectiveness of the method.