Towards testing big data analytics software: the essential role of metamorphic testing
Towards testing big data analytics software: the essential role of metamorphic testing
复制标题
DOI：
10.1007/s12551-018-0492-6
复制
发表时间：
2018-12
影响因子：
--
通讯作者：
Zhiyi Zhang;Xiaoyuan Xie
中科院分区：
文献类型：
--
作者：
Zhiyi Zhang;Xiaoyuan Xie
在快速发展的大数据分析领域，来自计算机科学和生物学等众多领域的科学家不断受到前所未有的数据量的挑战。虽然已经构建了许多软件程序来支持处理和分析连续信息流，但该领域的一个未被充分认识的挑战是这些大数据软件平台的软件质量保证。变形测试是为了缓解软件工程界的预言问题而提出的，它已经成为软件验证和确认的一种有效方法。近年来，我们见证了变形测试在从生物信息学到深度学习等各个领域的成功应用。在这封信中，我们回顾了变形测试在大数据上的一些主要应用，并对未来研究中的挑战提出了展望。
In the rapidly growing field of big data analysis, scientists from numerous domains such as computer science and biology are constantly challenged by an unprecedented amount of data. While many software programs have been constructed to support processing and analyzing continuous information flow, one under-appreciated challenge in this field is software quality assurance of these big data software platforms. Metamorphic testing, which was proposed to alleviate the oracle problem in the software engineering community, has become an effective approach for software verification and validation. Recent years, we have witnessed successful applications of metamorphic testing in a variety of domains, ranging from bioinformatics to deep learning. In this letter, we review some main applications of metamorphic testing on big data and present visions for the challenges in future research.