Abstract
This paper proposes several improvements on the conventional software reliability growth models (SRGMs) to describe actual software development process by eliminating an unrealistic assumption that detected errors are immediately corrected. A key part of the proposed models is the "delay-effect factor", which measures the expected time lag in correcting the detected faults during software development. To establish the proposed model, we first determine the delay-effect factor to be included in the actual correction process. For the conventional SRGMs, the delay-effect factor is basically non-decreasing. This means that the delayed effect becomes more significant as time moves forward. Since this phenomenon may not be reasonable for some applications, we adopt a bell-shaped curve to reflect human learning process in our proposed model Experiments on a real data set for internet-distributed software has been performed, and the results show that the proposed new model gives better performance in estimating the number of initial faults than previous approaches.