4. Increasing failure rate software reliability models for agile projects: a comparative study
Gabriel Ricardo Pena · 2020
A new software reliability model is proposed. The model is inspired in the Polya stochastic process, which is the asymptotic limit of the Polya urn model for contagion that is well described as a pure birth process. Modeling software reliability with this type of process introduces failure rate functions that depend not only on time but also on the number of failures previously detected. Since the failure rate function of the Polya stochastic process results in a linear-over-time mean number of failures, we propose a new pure birth process with a different failure rate function. This proposal results in a nonlinear-over-time mean number of failures and allows to model not only cases with increasing in time failure rate but also a reliability growth. We consider that failure reports collected under modern software engineering methodologies will require not just reliability growth models but also those which take into account increasing failure rate cases, if they are to be applied at the very first stage of development and testing, when code is constantly added either to fix failures or to accomplish new requirements. We show applications of our proposed model to several datasets and compare the performance with nonhomogeneous Poisson process models.