Abstract:

Constructing a software reliability model is an important method to evaluate the software reliability. The existing software reliability models usually assume that software faults are removed immediately when detected and new faults can not be introduced during the detection process, which is different from the real situations. To improve the above hypothesis conditions, a new software reliability growth model is presented. The detection process and the removing process are modeled as two independent non-homogenous Poisson processes in the new model. And the number of existing software faults is modeled as a changed function, including the initial faults and the newly introduced faults. Compared with other models, the presented model has better fitting ability and predicting ability by the fitting and predicting experiments with the two data sets.