| This dissertation aims to further advance the methodology of software reliability assessment based on Bayesian approach. In the proposed software reliability models, times between failures follow Weibull distributions with stochastically decreasing hazard rates reflecting the tester's intent to improve the software quality when carried out his corrective action.; Almost all technologies existing today rely heavily on computer and computer software. However, none of the existed software products is fault-free. Software errors have known to cause spectacular and sometimes catastrophic failures. Thus, from both developers and users' perspective, there is a great need for high quality software products. To improve software performance, major portion of resources such as money and time has been allocated for testing since testing can remove potential faults. However, during the testing process, almost all existing software reliability models assume only duration of testing, number of remaining faults, or times between failures affect software quality. In reality, software development is a complex process, and many extraneous environmental factors such as programmer skills, portion of added or changed codes, etc. influence the software reliability. Therefore, these factors need to be taken into consideration.; In this dissertation, based on Bayes approach, Weibull software reliability model incorporating a selected environmental factor, number of new or changed codes added during testing, is proposed to enhance the accuracy of software reliability assessment. To make the model more realistic, other practical issue such as error removal time is considered. Stop-test decision based on current software reliability, period of failure-free execution, and costs such as error removal cost, risk cost, testing cost are contemplated. The proposed models are validated using software failure data from real software applications. |
