Efficient reliability growth modelling for industrial software failure data

摘要

In this paper, we present a pragmatic approach for using stochastic reliability growth models in industrial software quality management. Key concept is the use of efficient algorithms for selecting suitable reliability growth models based on the evaluation of the model's appropriateness for long term prognoses. Compared to the generally accepted model evaluation criteria U-Plot, Prequential Likelihood and Holdout, the newly developed algorithms reduce speed and allow a selection of the model to be applied with regard to the kind of prognoses it will be used for. The significant increase of speed has shown to be crucial for application in industrial software projects. Our approach has been realised as an efficient, easy to use tool.