New interval Bayesian models for software reliability based on non-homogeneous Poisson processes

Abstract

We propose a new class of models for software reliability based on known models employing non-homogeneous Poisson processes, e.g., Musa-Okomoto and Goel-Okomoto models. We show that the general idea of model design is in a combined application of imprecise Bayesian inference and the maximum likelihood approach. We show examples where proposed models show better reliability prediction quality compared to the known ones.