A New Cascade Software Reliability Model

There are many kinds of software reliability models at present. However, a universal model, which can be used in different types of reliability data set, is never existed until now. This phenomenon blocks the development of software reliability model. In order to resolve the problem on software reliability model´s limitation and precision of model´s prediction, this paper proposed a new kind of method, with which four classical models are combined with a feedback neural network to form a new model. With the real world software failure data set SYSl and CSRl, we experimental investigated the new model and compared with four classical ones. It can be proved that the new model´s performance to predict is much better than any one of four classical models, and what´s more, it can be more widely used.