New interval bayesian software reliability models on the basis of the non-homogeneous Poisson processes

A new class of software reliability growth models is proposed. It is based on the well-known models using the non-homogeneous Poisson processes, for instance, Goel-Okumoto model or Musa–Okumoto model. The main idea of the models is to combine imprecise Bayesian models, where a set of prior probability distributions is considered instead of a single distribution. The numerical analysis of the proposed models with use of real statistical data indicates situations when the models provide higher reliability forecast quality in comparison with the known reliability models.