Regression Testing Selection and Dependency Analysis: A Design Science Study

Abstract

Continuous Large Scale Software Systems (LS3) development depends on Regression Testing Selection (RTS) to uphold quality of the software in a cost-effective manner. Prioritization-, minimization- and selection techniques are examples of this. We explore an existing technique and then classify it as a selection technique. We design and develop a new artefact, with the purpose of comparing the two Regression Testing Selection Techniques (RTST) with a select number of metrics. We compare the RTSTs using 500 commits with a one-tailed t-test which shows non-significant results. The artefact shows lower standard deviation for all metrics compared to the existing technique. The artefact identified, and removed 151 false positives that the existing technique would perform dependency analysis on. We conclude that the artefact shows improvements over the existing technique, while the random technique performed the worst.