The Application Method Research On Software Reliablity Models

As an important part of the software reliability engineering, the software reliability modeling has been researched on since 1970s. And until now, hundreds of software reliability models have been proposed. In order to facilitate model implementation, several software reliablity model tools have been developed. However, each of the model tools has some disadvantages such as limitation on application range and defect in human operation. In this dissertation, to overcome these shortcomings, the research work is summarized as follows:(1) Collect and classify the existing software reliability models, and analyze advantages and disadvantages of a few classic model classification methods.(2) Research on the parameter estimation of a widely used GO model and propose a weighted minimum least square method for it, which suggests that every data point should be assigned a weight according to its contribution of the prediction.(3) To be more pratical, the thesis proposes a user-oriented model selection strategy, which allows users to select models more efficiently.(4) Propose a method based on experiences from history projects, which considers the imformation of used models in history projects.(5) Design and implement a prototype system of software reliability prediction, which collects plenty of software reliability models and can be implemented on almost all the stages of software development, and apply the model selection methods proposed in this dissertation to make the model selection more humanity and diversified.