A stochastic software reliability growth model with Gompertz curve

In order to evaluate the quality level of software systems during the testing phase, deterministic statistical models such as Gompertz and logistic growth curve models have been used in many Japanese computer manufacturers and software houses. In this paper, the Gompertz growth curve model is reformulated as a stochastic model based on a nonhomogeneous Poisson process (NHPP). That is, introducing counting process representing the number of software errors detected up to testing time t, it adopts the modified Gompertz growth curve in the mean value function of the NHPP, which means the expected number of software errors detected during , where represents the probability of event and k= the expected error content in the system before testing, a, b= constant parameters. Then, several software reliability measures for quantitative assessment such as the expected number of software errors remaining in the system, the software reliability function, and the MTBF (mean time between software failures) can be derived. Numerical illustrations of software reliability measurement and assessment for a set of actual software error data based on the reformulated model are presented. [In Japanese.]