Confidence interval estimation of software reliability growth models derived from stochastic differential equations

Abstract

The study develops a software reliability growth model (SRGM) with confidence intervals that provides software developers useful information to decide the optimal software release time and to refine the quality of software testing tasks. The developed SRGM of this study is based on stochastic calculus to deduce the confidence intervals of the mean value function at different confidence level. Owing to less clear explanation of the variance in the mean value function of cumulative software errors in most software reliability growth models, it might not be effective in deducing the confidence interval concerning the mean value function. Therefore, software developers cannot estimate the possible risk variation in software reliability, and it might diminish the value of practical applications. In this study, we utilize the method of stochastic differential equations and five classic models (Goel and Okumoto's model (1979), Yamada's Delayed S-shaped model (1983), Ohba's Inflection S-shaped model (1984), Yamada's exponential model (1992), Chiu and Haung's learning effect model (2008)) to build the SRGM with confidence intervals that can assist the software developers in determining the optimal release times at different confidence levels. With regard to the software failure phenomena, they were supposed as Non-homogeneous Poisson Process (NHPP) in this study.