A hidden-Markovian software reliability model considering imperfect debugging and simultaneous removal of multiple faults

A new software reliability assessment model considering one or two simultaneous fault-removal activities and imperfect debugging environment is discussed in this paper. There have been developed various software reliability growth models (usually abbreviated as SRGMs) until now, and we are interested in the generalization of such models. Among the SRGMs proposed so far, although nonhomogeneous Poisson process models are widely-known as simple and useful ones, it is also known that they have several severe assumptions. That is, one of the assumptions is that just only one software fault can be removed at each debugging activity during the testing phase. This assumption contributes to simplify the model structure, however it has no guarantee whether such a simple model can describe the actual software debugging situations. Hence we propose a parametric model for assessing the software reliability during its testing phase as a hidden-Markovian model (HMM). In order to overcome the difficulty of the parameter estimation, it is known that the Baum-Welch re-estimation procedure is efficient. Therefore our model can estimate the probabilities of one or two simultaneous fault removal and imperfect debugging, even if a set of the software fault-removal time data is only available. This is the main advantage of our model. After deriving several software reliability assessment measures, we show software reliability analysis by using the actual data, and evaluate the goodness-of-fit of the model.