An investigation into whether the NHPP framework is suitable for software reliability prediction and estimation

Many software reliability growth models (SRGMs) based on non-homogeneous Poisson process (NHPP) framework have been proposed for estimating the reliability growth of products. However, some concerns regarding the properties of NHPP framework were exposed and discussed while the NHPP models have received considerable attention. Two main concerns are (I) the variance of an NHPP-based model grows as software testing proceeds, which was considered an unreasonable NHPP property for describing software failure behavior; and (II) the numbers of failures observed in disjoint time intervals are independent, which may fails in the early stage of software testing. With regard to Concern (I), we will justify the validity of NHPP framework through a mathematical perspective, i.e. the process of parameter estimation for NHPP models. Considering Concern (II), we will explain why NHPP SRGMs are still workable from the applicable perspectives. As a result, we believe the NHPP framework may still have merit.