Bayesian predictive analyses for non-homogeneous poisson process in software reliability with Musa-okumoto intensity function

Abstract

Due to rapid increase in development of complex computer systems over the past decades, there is need to estimate and predict the reliability of software systems during the testing process. Reliability refers to how well software meets its requirements and the probability of failure free operation for the specified period of time in a specified environment. The high demand and use of software has led to increased quest for more reliable software. For the past few decades several software reliability growth models have been used to describe the behavior of software testing process. Predictive analyses of software reliability model is of great importance for modifying, debugging and determining when to terminate software development testing process. This study performed one-sample Bayesian predictive analyses for Musa – Okumoto software reliability model using informative and non – informative priors. The study mainly focused on four issues on single-sample case that have been outlined in chapter three as issue A, B, C and D that relate to software development testing process. Simulated and secondary data were used to illustrate these issues. For secondary data, Goodness of Fit (GOF) test based on Laplace statistics was performed to check whether the model fit well to the data before it was used to illustrate the derived methodologies and It was found fit well to the data. The study developed explicit solutions to the issues and on issue D coverage probability was computed and found to be a good estimator and thus it will help to solve problems related to reliability of developed software and make a trade-off decision in software industry.