Abstract
In this paper, a new model selection method, software reliability model selection method based on unascertained set, is proposed to remove the inconsistency of the software reliability model application. The new method starts from a single index value, in which the unascertained set theory is adopted to comprehensively evaluate all the indexes, entropy is used to compute the classification weight of indexes, identification criteria of the degree of confidence are applied to identify the final model. An example is presented to show the procedure and the reasonable result of the new method. The software reliability model selection method based on unascertained set can effectively reduce the inconsistency of the reliability models. Its application prospect is extensive.