Abstract
In this paper, unascertained mathematics theory is applied to software reliability modeling research. It is used to study on the software fault process, and to discover software failure mechanics. The failure data which are used for prediction are amended according to the growth characteristics of software. On this foundation, a software reliability model, based on unascertained rational number, is proposed. The modeling process of the new model is based on a set of assumptions, and is different from the statistical distribution supposition of various s laws about the failure rate in the traditional software reliability modeling process. Therefore, it has better applicability in theory. We select 4 groups of engineering data, analyze and compare separately the new model with J-M model and G-O model. The results testify that the new model has better applicability, which can improve the incongruent of model application efficiently to some extent.