Research Of Software Reliability Assessment Technology Based On Bayesian Network

Software reliability occupies an important place in computer engineering. Evaluation of software reliability, which is an important content of software reliability research, can quantize Parameters of software reliability and make software more reliable. Software reliability model, as an important tool of evaluation, forecast and analysis, is the core of software reliability research. These existing models describe software reliability to some extent. However, because of harsh conditions or having not utilizing prior information sufficiently, these models are not pervasively applicable.Bayesian Network has characteristics such as graphic representation of modeling methods and ability of performing incomplete and uncertain data reasoning, which has distinct advancement compared with traditional modeling methods. The thesis use Bayesian Network in the software reliability assessment, the results are as follows:(1) Propose the improvement software reliability predication model based on Bayesian network, which uses the prior information and considers the incomplete error correction and error correction time, improves the prediction accuracy and validity;(2) Propose software operational profile model based on the Bayesian Network, which fully uses prior information, and could reduce the estimation error, makes the operational profile conform to the reality;(3) Propose software reliability assessment model based on Bayesian Network, which takes the software operational profile as foundation, and fully considers the software reliable correlation influence factors. The model may realize the dynamic evolution and advancement, and estimate reliability degree more precisely and stably;(4)Construct a software reliability assessment prototype system, and confirm method of the thesis to be effective.Finally, the research work involved in the thesis is summarized and the future developments in RBF network learning are forecasted.