This paper presents a tertiary review of software quality measurement research. To conduct this review, we examined an initial dataset of 7,811 articles and found 75 relevant and high-quality secondary analyses of software quality research. Synthesizing this body of work, we offer an overview of perspectives, measurement approaches, and trends. We identify five distinct perspectives that conceptualize quality as heuristic, as maintainability, as a holistic concept, as structural features of software, and as dependability. We also identify three key challenges. First, we find widespread evidence of validity questions with common measures. Second, we observe the application of machine learning methods without adequate evaluation. Third, we observe the use of aging datasets. Finally, from these observations, we sketch a path toward a theoretical framework that will allow software engineering researchers to systematically confront these weaknesses while remaining grounded in the experiences of developers and the real world in which code is ultimately deployed.
翻译：本文介绍了对软件质量衡量研究的三级审查。 为了进行这一审查,我们审查了7 811篇文章的初步数据集,并发现了75项与软件质量研究相关和高质量的二级分析。我们综合了这一方面的工作,概述了各种观点、计量方法和趋势。我们从五个不同的角度确定了质量概念,将质量概念分为:超能力、可维持性、整体概念、软件的结构特征和可靠性。我们还确定了三个关键挑战。首先,我们发现了具有共同措施的关于有效性问题的广泛证据。第二,我们观察了机器学习方法的应用,但没有进行充分的评估。第三,我们观察了老化数据集的使用情况。最后,我们从这些观察中勾画出一条通往理论框架的道路,使软件工程研究人员能够系统应对这些弱点,同时继续以开发者的经验和最终部署代码的真实世界为基础。
相关内容
微信扫码咨询专知VIP会员