MS-guided many-objective evolutionary optimisation for test suite minimisation

作     者：Zheng, Wei Wu, Xiaoxue Cao, Shichao Lin, Jun 

作者机构：Northwestern Polytech Univ Sch Software & Microelect 127 West Youyi Rd Xian Shaanxi Peoples R China Northwestern Polytech Univ Sch Automat 127 West Youyi Rd Xian Shaanxi Peoples R China 

出 版 物：《IET SOFTWARE》 (IET软件)

年 卷 期：2018年第12卷第6期

页      面：547-554页

核心收录：

学科分类：0808[工学-电气工程] 08[工学] 0835[工学-软件工程] 

主　　题：statistical analysis Pareto optimisation minimisation genetic algorithms program testing statistical analysis nondominated sorting genetic algorithm II program space software-artefact infrastructure repository MS-guided many-objective optimisation approach optimal test suite test cost NSGA-II test suite minimisation process standard code coverage criteria mutation score-guided many-objective optimisation approach many-objective optimisation problem redundant test cases obsolete test cases MS-guided many-objective evolutionary optimisation 

摘      要：Test suite minimisation is a process that seeks to identify and then eliminate the obsolete or redundant test cases from the test suite. It is a trade-off between cost and other value criteria and is appropriate to be described as a many-objective optimisation problem. This study introduces a mutation score (MS)-guided many-objective optimisation approach, which prioritises the fault detection ability of test cases and takes MS, cost and three standard code coverage criteria as objectives for the test suite minimisation process. They use six classical evolutionary many-objective optimisation algorithms to identify efficient test suite, and select three small programs from the Software-Artefact Infrastructure Repository (SIR) and two larger program space and gzip for experimental evaluation as well as statistical analysis. The experiment results of the three small programs show non-dominated sorting genetic algorithm II (NSGA-II) with tuning was the most effective approach. However, MOEA/D-PBI and MOEA/D-WS outperform NSGA-II in the cases of two large programs. On the other hand, the test cost of the optimal test suite obtained by their proposed MS-guided many-objective optimisation approach is much lower than the one without it in most situation for both small programs and large programs.