基于退火Memetic算法和贝叶斯网络的回归测试用例集优化研究

项目名称： 基于退火Memetic算法和贝叶斯网络的回归测试用例集优化研究

项目编号： No.61202030

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 计算机科学学科

项目作者： 王赞

作者单位： 天津大学

项目金额： 22万元

中文摘要： 回归测试用例约简及优先排序技术作为可降低回归测试成本的有效手段，已被广泛用于回归测试中。遗传算法等被应用于求解约简问题并取得了一定的效果，收敛速度慢及过早收敛等问题使得算法还有一定的改进空间。基于覆盖率等技术对测试用例优先排序问题的研究也取得了一定的进展，但以往方法多单独考虑影响排序的因素，如综合考虑将可能产生更为可靠的排序标准。本课题首先将开发一个回归测试综合信息平台，对测试的静态及动态信息进行记录，为优化提供基础；其次，为改善遗传算法相关问题，拟将模拟退火和Memetic的混合算法应用于求解用例集约简问题；再次，综合考虑影响用例优先排序的因素，特别是检测到"重要错误"的因素，构建评价测试用例检测到"重要错误"概率的贝叶斯网络，为用例集优先排序提供重要的排序标准；最后，综合考虑测试优化的三个阶段，构建一套包括"选择-约简-排序"的回归测试优化平台。本课题将通过真实数据对所提算法进行比较。

中文关键词： 软件测试；错误易发性模型；贝叶斯网络；进化算法；软件缺陷定位

英文摘要： As effective methods to reduce testing costs, regression test case reduction and test case prioritization techniques have been widely hired to regression test. Genetic algorithm and other means are employed to solve problems related to test case reduction and some effects are achieved, whereas the algorithm still needs to be improved due to its slow convergence speed and premature convergence to local optima. Researches on test case prioritization problems based on coverage and other techniques also made great development. However, most previous methods consider factors influencing prioritizing process separately. A more reliable sorting criterion may be produced under integrated consideration. This research first develops a comprehensive information system for regression test, which records static and dynamic information in the process of testing. To improve the genetic algorithm, hybrid algorithm based on simulated annealing and memetic algorithm is employed to solve test suite reduction problem. Thirdly, after investigating factors influencing test case prioritization, especially factors detecting important errors, this research builds a Bayesian network which can evaluate probability of test case detection of important errors, in view of providing key criterion to test suite prioritization. Finally, consider

英文关键词： Software Testing；Error Proneness model；Bayesian Network；Evolutionary Algorithm；Software Fault Localization