Software reliability prediction based on attributes data in design review and testing processes

It is an important problem to achieve high software reliability for a computer system to support high level of our information society. Considering the improvement of software productivity, it is necessary to predict and estimate software reliability in terms of latent software faults in the early development phase. Generally, the software development process is divided into design, coding, testing, and maintenance phases. The review activity is performed to check. the quality of intermediate software products and detect/remove faults introduced in them as quality control technology rafter each phase. Especially, as generally known, the design review is a very effective approach to improving software productivity and quality. This paper proposes software reliability prediction models based on the attributes data in the design review and testing processes. The models proposed here are compared with the existing model by using the data obtained from actual software projects. The result of actual experiments shows, that the proposed models considering the attributes such as review-case density, fault-detection rate in design review, test-case density, and fault-detection rate in testing are useful for improving the productivity and reliability of intermediate and final software products.