Bayesian Networks applied to Software IV&V

In this paper, we describe how Bayesian networks can be used to merge quantitative and qualitative information to support IV&V of use cases. Essentially, simple metrics from the use cases are computed, which are then input to a Bayesian network. This network models the relationships between the observable parameters of an IV&V process for use cases, and the desired features of the requirements specifications. The output of the network is an assessment of the maturity of the requirements, in terms of the probability that they exhibit the desired properties. We apply our proposed approach to a real system: a software simulator built to test attitude control for an aerospace system, to illustrate how IV&V can be quantitatively supported