A study on software reliability prediction based on triple exponential smoothing method (WIP)

In the software testing process, the effective development of software reliability models can be quantified to assess the software reliability. Most of the current software reliability models developed are the parametric models, and assuming that the software fault detection process is Markov or non-homogeneous Poisson process (NHPP). However, due to the complexity of the software testing process, assumptions of building the models can not satisfy the actual software testing situation, and those parametric models set up by assumptions can not accurately predict the software failure process. In this case, we propose a non-parametric method of the software reliability prediction based on the triple exponential smoothing. We compare the proposed non-parametric method with the double exponential smoothing and other software reliability model. The experimental results show that using the proposed non-parametric method can effectively and accurately predict the number of software failure. Furthermore, it does not need a lot of historical fault data and calculations, and can be easily made tool used in the actual software test.