Bayesian concepts in software testing: an initial review

Abstract

This work summarizes the main topics that have been researched in the area of software testing under the umbrella of ``Bayesian approaches'' since 2010. There is a growing trend on the use of the so-called Bayesian statistics and Bayesian concepts in general and software testing in particular. Following a Systematic Literature Review protocol using the main digital libraries and repositories, we selected around 40 references applying Bayesian approaches in the field of software testing since 2010. Those references summarise the current state of the art and foster better focused research. So far, the main observed use of the Bayesian concepts in the software testing field is through the application of Bayesian networks for software reliability and defect prediction (the latter is mainly based on static software metrics and Bayesian classifiers). Other areas of application are software estimation and test data generation. There are areas not fully explored beyond the basic Bayesian approaches, such as influence diagrams and dynamic networks.