Toward a K-means clustering approach to adaptive random testing for object-oriented software
Toward a K-means clustering approach to adaptive random testing for object-oriented software
复制标题
面向对象软件的自适应随机测试的 K 均值聚类方法
DOI：
10.1007/s11432-018-9827-9
复制
发表时间：
2019-09
期刊：
影响因子：
--
通讯作者：
Chengying Mao
中科院分区：
文献类型：
--
作者：
Jinfu Chen;Minmin Zhou;T.H. Tse;Tsong Yueh Chen;Yuchi Guo;Rubing Huang;Chengying Mao
1江苏大学计算机科学与通信工程学院，镇江212000，中国；2江苏大学工业网络空间安全技术江苏省重点实验室，镇江212000，中国；3计算机科学系
Testing and debugging are mainstream methods for software quality assurance. In particular, random testing (RT, also known as fuzz testing) and partition testing (PT) are most widely adopted. Classical studies [1, 2] show that PT is only marginally better than RT in many cases but has considerably more overheads. On the other hand, RT does not consider the fact that failure-causing inputs tend to be conglomerated into regions. To address this issue, Chen et al.[3, 4] proposed adaptive random testing (ART), which targets at spreading the test cases as evenly as possible across the entire input domain. The failuredetection effectiveness is improved. Object-oriented (OO) software has become the de facto standard in the industry. However, traditional testing methods are not immediately applicable. ART is no exception. Adaptations are needed. Ciupa et al.[5] presented the notion of object distance and proposed ART for objectoriented software (ARTOO) based on the original fixed-sized-candidate-set ART (FSCS-ART) algorithm [3]. Chen et al.[6] further proposed the object and method invocation sequence similarity (OMISS) metric for OO software, covering not only distances between objects but also between method invocation sequences. Their OMISSART algorithm is based on FSCS-ART as well as the max-min criterion and the forgetting strategy.In particular, the forgetting strategy [7] simply considers some of the already executed test cases when selecting the next test case, thus reducing the computational overhead. In this letter, we propose an alternative approach to the selection of next test cases for OO software, applying the concept of K-means clustering [8]. Cluster analysis gathers similar data together, and has been widely used in many fields. Thus, we group test cases with similar properties into the same cluster. In other words, objects or method invocation sequences (MIS) within a cluster will be similar to one another, and objects or MIS in different clusters will be dissimilar. We then compare any new candidate with all the already executed test cases collectively, thus simplifying the test case selection process.
登录
查看更多内容
DOI：
10.1007/978-3-540-30502-6_23
发表时间：
2004-12
期刊：
--
影响因子：
--
作者：
Tsong Yueh Chen;H. Leung;I. K. Mak
通讯作者：
Tsong Yueh Chen;H. Leung;I. K. Mak
DOI：
10.1109/wst.1988.5376
发表时间：
1988-07
期刊：
[1988] Proceedings. Second Workshop on Software Testing, Verification, and Analysis
影响因子：
--
作者：
R. Hamlet;R. Taylor
通讯作者：
R. Hamlet;R. Taylor
影响因子：
4.3
作者：
Stankovic, RS;Falkowski, BJ
通讯作者：
Falkowski, BJ
影响因子：
7.4
作者：
HAMLET, D;TAYLOR, R
通讯作者：
TAYLOR, R
影响因子：
7.4
作者：
DURAN, JW;NTAFOS, SC
通讯作者：
NTAFOS, SC