SOFTWARE RISK ANALYSIS
Norman F. Schneidewind · International Journal of Reliability Quality and Safety Engineering · 2009
There has been a lack of attention to the subject of risk management in the design and operation of software. This is strange because the risk to reliability is a critical problem in attempts to achieve a safe operation of the software. To address this problem, we evaluate existing models and introduce a new model for software risk prediction. The new model — cumulative failures gradient function — is based on the principles of neural networks. This metric identifiers the minimum test time required to achieve maximum improvement in software quality. We used three NASA Space Shuttle software systems in the evaluation of both existing and new models. The results showed that it was not possible to consistently rank these systems because the validity of the risk predictions varied depending on the risk model that was used. Therefore, the results suggest that it is advisable to use a variety of models to comprehensively evaluate the software risk.