Non-homogeneous Inverse Gaussian Software Reliability Models

In this paper we consider a novel software reliability modeling framework based on non-homogeneous inverse Gaussian processes (NHIGPs). Although these models are derived in a different way from the well-known non-homogeneous Poisson processes (NHPPs), they can be regarded as interesting stochastic point processes with both arbitrary time non-stationary properties and the inverse Gaussian probability law. In numerical examples with two real software fault data, it is shown that the NHIGP-based software reliability models could outperform the goodness-of-fit and the predictive performances more than the existing NHPP-based models.