Improving Software Regression Testing Using a Machine Learning-Based Method for Test Type Selection
Paper i proceeding, 2022

Since only a limited time is available for performing software regression testing, a subset of crucial test cases from the test suites has to be selected for execution. In this paper, we introduce a method that uses the relation between types of code changes and regression tests to select test types that require execution. We work closely with a large power supply company to develop and evaluate the method and measure the total regression testing time taken by our method and its effectiveness in selecting the most relevant test types. The results show that the method reduces the total regression time by an average of 18,33% when compared with the approach used by our industrial partner. The results also show that using a medium window size in the method configuration results in an improved recall rate from 61,11% to 83,33%, but not in considerable time reduction of testing. We conclude that our method can potentially be used to steer the testing effort at software development companies by guiding testers into which regression test types are essential for execution.

Machine learning

Regression test types

Software regression testing

Code types

Författare

Khaled Al Sabbagh

Göteborgs universitet

Miroslaw Staron

Göteborgs universitet

Regina Hebig

Göteborgs universitet

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

03029743 (ISSN) 16113349 (eISSN)

Vol. 13709 LNCS 480-496
9783031213878 (ISBN)

23rd International Conference on Product-Focused Software Process Improvement, PROFES 2022
Jyväskylä, Finland,

Ämneskategorier

Programvaruteknik

Datavetenskap (datalogi)

Datorsystem

DOI

10.1007/978-3-031-21388-5_33

Mer information

Senast uppdaterat

2023-10-27