Parameter Estimation of Software Reliability Growth Model Using Metropolis-Hastings Algorithm and Release Policy
Da Hye Lee, In-Hong Chang · Journal of Applied Reliability · 2024
Purpose: In this study, the Metropolis-Hastings (M–H) algorithm is used to estimate parameters of the software reliability growth model, and release policies are discussed. Methods: The parameters of the mean value function, as obtained using the M-H algorithm, are estimated by fitting actual failure datasets to the Goel–Okumoto model. Additionally, to formulate release policies, development costs and release time are examined based on changes in the coefficients of the cost model. Results: We estimate the Goel-Okumoto model by using the M-H algorithm. The obtained estimates are then substituted into the cost model to ascertain the effects of changes in certain cost coefficients on software development costs and release time. Conclusion: In conclusion, the M-H algorithm can be used as an alternative to MLE for estimating the parameters of the software reliability growth model. Because the prior distribution of the parameters of the SRGM is not known, various distributions must be considered as proposal distributions.