Software reliability modelling and prediction with hidden Markov chains

The purpose of this paper is to use the framework of hidden Markov chains (HMCs) for the modelling of the failure and debugging process of software, and the prediction of software reliability. The model parameters are estimated using the forward-backward expectation maximization algorithm, and model selection is done with the Bayesian information criterion. The advantages and drawbacks of this approach, with respect to usual modelling, are analysed. Comparison is also done on real software failure data. The main contribution of HMC modelling is that it highlights the existence of homogeneous periods in the debugging process, which allow one to identify major corrections or version updates. In terms of reliability predictions, the HMC model performs well, on average, with respect to usual models, especially when the reliability is not regularly growing.