Abstract: Software reliability growth models are used for the prediction of software reliability which can decided whether software could make release. At present, all kinds of common models' predict ability is inconsistent. We proposed a framework for software reliability growth model selection and application. The method provided guidelines on how to select better model set on the given failure data set, and then used these models to predict reliability with neural network. I}he method was applied in a case study using failure data from some real world software projects and the experiment indicates its efficiency.

Key words: Software reliability, Software Reliability Growth Model, Model selection and compose, Neural network