Enhancing the accuracy of software reliability prediction through quantifying the effect of test phase transitions

摘要

Since marketing unreliable software products will lead to customer dissatisfaction, the products usually undergo several phases of testing to minimize the number of faults before being released to the market. However, because of budget limitations and time constraints, very long test periods are impractical. Thus, project managers need to balance the cost of testing and the possible effects of any remaining (undetected) faults. Software reliability growth models (SRGMs) can be used to predict the fault detection process and help project managers determine a cost-effective time to stop testing and release the product. In practice, software testing may be divided into several phases, each of which has a different objective. Few existing SRGMs consider the influence of test phase transitions even though they may have a significant effect on fault detection during the test phase. Therefore, to address this research gap, we quantify the variations in the effect of different test phases and propose a software reliability modeling framework. The SRGMs obtained from the proposed framework can be used to gauge the influence of test phase transitions. We validated the framework’s performance on a failure data set collected from a real software project. The results demonstrate that the proposed framework accurately reflects the influence of test phase transitions and yields a strong performance in terms of fitting as well as predicting the fault detection process.