一种基于规则提取的自动化测试用例生成方法

An Effective Algorithm for Automated Test Suite Generation Based on ANN(Artificial Neural Netwrok) Rule-Extraction

软件测试过程中,缩小测试用例集的规模,通常需要根据经验进行分析,找出输入输出间的非映射关系,这往往要耗费过多的测试资源。而人工神经网络在此问题的处理上有其特有的优势。因此文章提出了一种改进的规则提取方法,用于生成测试用例。通过构建神经网络模型,建立输入/输出之间的非线性映射关系,接着根据连接的权值,裁剪网络,去除与特定输出无关的输入属性。然后,在规则提取阶段仅保留两个与该输出最为相关的输入,并由此提取出IF-THEN规则,生成测试用例。文章完成了改进后规则提取算法各阶段的自动化,显著降低了在测试用例设计环节上的开销。最后,通过程序验证了该方法的有效性。

Sections 1 and 2 of the full paper explain the algorithm mentioned in the title,which we believe is more effective than previous ones.Their core consists of：＂Test suite minimization usually needs analyzing the nonlinear relationships between input and output,which is an expensive,tedious and error-prone process in software testing.ANN has its unique advantages on these issues.We propose an approach that uses an improved ANN rule extraction method to generate test cases.For the specified output,this algortthm removes less significant inputs according to their weight magnitudes.There will be only at most two inputs according to their specified output.Then,we extract IF-THEN rules from a pruned ANN,which can be used to generate test cases.And all these pruning phase and rule extraction phase can be automated.＂Section 3 uses a small program to prove its validity;after analysis,section 3 presents the results in Table 2,thus showing that our algorithm reduces indeed considerably test domain analysis.