Optimization selection method for software reliability growth model based on cosine similarity

So far, there have been many different software reliability growth models (SRGMs) established. It is difficult to choose which SRGM to apply in the reliability evaluation of actual software projects due to the varying assumptions of the established SRGM. In general, there will be considerable discrepancies between SRGMs used for fault prediction and software reliability evaluation of the same software development project. Considering the complexity of the actual software testing process, selecting a single optimal SRGM to evaluate software reliability may not be in line with the actual situation of fault detection (FD) or fault introduction (FI) during software testing. In order to select a class of appropriate SRGMs for the current software development and testing environment in the actual software project reliability evaluation, this paper proposes using the cosine similarity classification method. The purpose of this study is to explore effective methods for dividing into a class of optimal models, rather than selecting an optimal model. In comparison to the classical distance based approach (DBA) for selecting a single optimal SRGM, the proposed method can effectively partition a class of optimal models, including a single optimal model selected by DBA. Experimental results demonstrate the effectiveness of the proposed method.