A Discussion of Software Reliability Growth Models with Time-Varying Learning Effects

Abstract

Over the last few decades, software reliability growth models (SRGM) has been developed to predict software reliability in the testing/debugging phase. Most of the models are based on the Non-Homogeneous Poisson Process (NHPP), and an S or exponential-shaped type of testing behavior is usually assumed. Chiu et al. (2008) provided an SRGM that considers learning effects, which is able to reasonably describe the S and exponential-shaped behaviors simultaneously. This paper considers both linear and exponential-learning effects in an SRGM to enhance the model in Chiu et al. (2008), assumes the learning effects depend on the testing-time, and discusses when and what learning effects would occur in the software development process. This research also verifies the effectiveness of the proposed models with R square (Rsq), and compares the results with these of other models by using four real datasets. The proposed models consider constant, linear, and exponential-learning effects simultaneously. The results reveal the proposed models fit the data better than other models, and that the learning effects occur in the software testing process. The results are helpful for the software testing/debugging managers to master the schedule of the projects, the performance of the programmers, and the reliability of the software system.