To understand how software quality improves during testing, SRGMs play a significant role. SRGMs are even more critical, especially when human learning, debugging practices, and uncertainties in the testing environment influence fault detection processes. Traditional SRGMs often assume ideal or simplified conditions, which may not fully reflect real testing scenarios. To address this gap, this study proposes two SRGMs that consider imperfect debugging, S-shaped learning pattern, and uncertainty modeled through a Weibull distribution. These additions allow the models to capture fault detection behaviors more realistically. The performance of the proposed models is evaluated using three datasets and compared with some well-known SRGMs. Results show that the proposed models perform competitively across all datasets. Although no single model dominates across all cases, the results indicate that combining learning and uncertainty provides flexibility and enhanced predictive ability in various testing environments.
Samal et al. (Thu,) studied this question.