A BP neural network based hybrid model for software reliability prediction

To overcome the shortcomings of traditional software reliability models and adapt to the current software and its development process characterized by increasing complexity, this paper proposes a hybrid model with BP neural network model serving as a nonlinear hybrid system of several traditional models, to improve the prediction accuracy. Because of its input diversity, the model can be used universally. Experiment showed that this model not only retained the experience of traditional software reliability models, but also combined BP neural network good nonlinear mapping ability and good generalization.