Techniques Based on Data Science for Software Processes: A Systematic Literature Review

Replication data for: Qualities of quality: A tertiary review of software quality measurement research

“…With respect to validity, Syeed et al [4] found in their systematic review of contributions from software quality research to open source development practices that many of the most significant quality predictors were rarely used in research. Prior authors have also identified potential avenues for progress: mining software repositories unlocks longitudinal data for quality research [5], and AI/data science techniques have much to contribute if data and metric validity concerns can be addressed [6]. We found one previous tertiary analysis in the field of software quality: Elmidaoui et al [7] examined nine secondary studies of software maintainability as a specific facet of quality and found that while maintainability prediction is an active area of work, model performance and validation continue to be a concern.…”

“…Machine learning methods are increasingly being used to measure quality [6] and ML studies appear in our corpus from multiple perspectives [S5, S15, S20, S23, S25, S28, S62, S71]. However, several authors found that a lack of shared reporting standards for evidence and evaluation in ML is substantially hampering progress toward the use of ML to support evidence-driven engineering [S20, S23, S71].…”

“…Future researchers considering data science techniques for assessing software quality may find the clear reports made by Fernandez Del Carpio and Angarita [6] with respect to ML thresholds and model fitness to be of particular value. Addressing these reporting issues is relatively straightforward, but requires attention from researchers as well as reviewers and editors.…”