Statistical Estimation of Imperfect Debugging Rate Based on Hidden-Markov Software Reliability Modeling

摘要

This paper considers a simple software reliability assessment model in imperfect debugging environment, and proposes a method for estimating a parameter of imperfect debugging rate. Many stochastic models for software reliability assessment have been proposed so far, and some of them have been applied in the actual software testing management. Among these models, the so-called imperfect debugging models have some advantage in terms of the fitness to the actually observed data, because the assumption of perfect debugging is unrealistic in the actual software testing-phase. However, the imperfect debugging software reliability assessment models have not prepared the practical method for estimating the imperfect debugging rate itself. In other words, by using the method of maximum-likelihood, we can not estimate the parameter only from software failure data. We discuss the method of estimating the imperfect debugging rate by using the Baum-Welch re-estimation procedure in hidden-Markov modeling. Several numerical illustrations are shown and then we argue several estimation results.