Considering the testing process of the software system as a stochastic process is a primary approach to the software reliability modeling technique. Besides some popular distributions, the Poisson distribution has been considered the best based on its advantage when modeling the times at which arrivals enter a system. In the non-homogeneous Poisson process group of models, the S-shaped function is a value curve with many good results. This paper proposes a new imperfect debugging software reliability model based on (1) an Imperfect debugging assumption (the testing process could cause new faults); and (2) the Fault detection rate can be controlled more effectively by the appearance of a growth-rate-controller. The real data from industrial projects verify the application of this model based on good popular criteria values.

Abstract Software reliability modelling is the mathematical technique used to evaluate the reliability of a software system. The non‐homogeneous Poisson process is a prominent approach in this field. More than half of the models in this group are based on S‐shaped functions, primarily the 2‐parameter S‐shaped function. This paper proposes a new model based on the 3‐parameter S‐shaped function, which is an expanded form of the 2‐parameter S‐shaped function obtained by adding a growth rate controller. Real data from industrial software development projects are used to verify the usability of the proposed model. The proposed model is shown to perform better than the existing models, especially with respect to the predictive performance. Furthermore, the rate of convergence of the proposed model is acceptable, with a rate of 76.47%.