基于复杂网络的软件测试路径聚类分析-计算机工程与应用
72 2010 ,46 (31 ) Computer Engineering and Applications 计算机工程与应用
基于复杂网络的软件测试路径聚类分析
周宽久,冯金金,兰文辉,迟宗正
ZHOU Kuan-jiu ,FENG Jin-jin ,LAN Wen-hui ,CHI Zong-zheng
大连理工大学 软件学院,辽宁 大连 116620
Software School ,Dalian University of Technology ,Dalian ,Liaoning 116620,China
E-mail :fjj.dut@ 163.com
ZHOU Kuan-jiu ,FENG Jin-jin ,LAN Wen-hui ,et al.Cluster analysis of software testing paths based on complex net-
works.Computer Engineering and Applications ,2010 ,46 (31 ):72-76.
Abstract :A complex networks-based algorithm that can get fewer software testing paths is put forward to reduce the num-
ber of test data ,and to improve test efficiency.Firstly the method of calculating the nodes ’weight is advanced.This method
is based on the weighted complex networks model of software system.Then a similarity matrix is created according to the
uncertainty during software running.Finally ,a matrix decomposition algorithm is issued to convert the matrix to obtain the
partition of testing paths.After cluster analysis of the seven Linux files ,the test paths are on average 17.46% saves.It is
shown that this method is simple ,rapid and effective from both theoretical analysis and experiment.
Key words :complex networks ;software testing ;cluster analysis ;similarity matrix
摘 要:为了在保证覆盖率的前提下,减少测试用例数量,提高测试效率,提出一种基于复杂网络的软件测试路径生成算法。首
先提出软件系统加权复杂网络模型中节点权值的计算方法,建立了复杂网络模型,然后根据软件运行时的不确定性,建立函数调
用的模糊相似矩阵,并利用相似矩阵结构分解算法对其进行变换,对测试路径进行聚类分析得到测试路径的划分。分析Linux 的
7 个文件,经过聚类分析后平均节约了17.46%的测试路径。理论分析和实例证明,方法简单可行。
关键词:复杂网络;软件测试;聚类分析;相似矩阵
DOI :10.3778/j.issn. 1002-8331.2010.31.020 文章编号:1002-8331(2010 )31-0072-05 文献标识码:A 中图分类号:TP311
1 引言 网络研究的深入,人们发现大型软件系统中也存在复杂网络
[5-6]
随着软件系统越来越复杂,如何用尽量少的测试用例覆 特征 。同时,为了使得不同类中的数据尽可能相异,同一类
盖尽量多的测试路径成为当务之急。 中的数据尽可能相似,人们通常把一组组个体按照相似性归