A Bayesian Prediction Model for Risk-Based Test Selection

In industry, testing is commonly performed under severe pressure due to limited resources. Therefore, risk-based testing, which uses predicted risks to guide the test process, is employed to select test cases. To this end, risks have so far mainly been estimated ad hoc, but not systematically predicted on the basis of the defect history and defect costs. In this paper, we present a novel approach to risk-based test selection, which employs a comprehensive and versatile Bayes risk model taking defect probabilities and costs into account. It enables the prediction of a risk decrement that could potentially be used for test selection. We first define a generic Bayes risk decision criterion for test selection, and then implement and evaluate it in an industrial software development project, where it is intended to support decisions steering the quality assurance process.