Stochastic software reliability growth modeling based on discrete statistical data analysis models

摘要

A quantitative and precise assessment of software reliability in the testing phase is an important issue to provide a software system keeping high degree of reliability for the user because the testing phase is the final stage of software development. Up to now, as one of the methods to assess software reliability quantitatively, a number of SRGM's (software reliability growth models) have been proposed. Especially, NHPP (nonhomogeneous Poisson process) model is known as a useful one for assessing and forecasting software reliability of the developed software system. At present, the NHPP models contribute to software reliability assessment in many computer manufactures and software houses. On the other hand, there is a case that statistical data analysis models are adopted to assess software reliability in Japan. In recent years, discrete ones which enable us to assess more accurately have been proposed. In this paper, we develop stochastic software reliability growth models by revising the discrete statistical data analysis models. And we show results of goodness-of-fit comparisons with the discrete statistical data analysis models by using data sets observed in actual testing phase.