Using a cluster analysis method for grouping classes according to their inferred testability: An investigation of CK metrics, code coverage and mutation score

摘要

Software testing is one of the most time and resource consuming activities in every software development process. The effort to test a software is given by its testability, a quality indicator that can be indirectly measured to indicate how easily a software can be tested. Researchers have been studying the influence of object oriented metrics, such as the CK metrics, on the testability of a software. To do so, they correlate the CK metrics with the test size and also with test quality indicators, such as code coverage and mutation testing. However, those studies provide only evidence of the correlation between each metric but they do not provide any information regarding the range of values for which a reasonable amount of effort is spent or the quality of the test set considered. In this paper, we present an analysis in which we split the classes of four open source software into several clusters according to their CK metrics and show the ranges for which they have high or low testability according to their code coverage and mutation score.