An empirical study of regression testing techniques

Abstract

Regression testing is a testing process for established software and developed ones. Adding features to the software should not cause errors in the previous software version. There are several methods for implementing regression testing. The most straightforward is to do retest-all, but this technique is costly and protracted because it has to execute all test cases in the test suite. This situation is the nature of regression testing research, i.e. determining how to execute only a few test cases to shorten the time, and the faults can be found as much as possible. Recently there are various algorithms for regression testing process. This paper explains an empirical study of several regression testing methods. In the empirical evaluation, we used open-source SUTs, and several are implemented ourselves. On the other side, we utilize an MTS tool from the software artefact repository (SIR) and other open-source tools to automatic testing execution. The empirical study shows that each technique has a specific characteristic, and also has effectiveness significantly to reduce execution time testing process. So that in the regression testing implementation, it is crucial to determining the techniques considering each of them has consequences.