An approach for early prediction of software reliability

In the early stages of development, failure information is not available to quantitatively measure reliability of a software product. In this context, we propose an approach to predict software reliability early in the product development stages from design metrics. First we predict reliabilities of the components of a system. For this, we categorize the different kinds of faults that can occur in a component during its development and identify the design metrics that correlate to these faults. We construct a Bayesian Belief Network (BBN) model to predict reliabilities of the components using the identified design metrics. Based on predicted reliabilities and usage frequencies of the components of a system, we determine the reliability of the system. The applicability of our proposed model is illustrated through a case study. Results obtained from our case study indicate the effectiveness of our approach