A Model for Program Error Prediction Based on Software Reviewing Process Characteristics and Its Evaluation

摘要

At present, two types of software reliability models have been proposed. 1)～6) One is the so called dynamic reliability model, the other is the so called static reliability model. In order to resolve prediction difficulty of above two type of software reliability models, we have been already proposed new software reliability models based on testing characteristics.7) In such a model, the test case density and error detection rate are easily measured after the end of testing. This model is useful to a ssess the possibility of software product release. Considering about the improvement of software productivity, it is necessary to predict and evaluate software reliability more quickly, and to perform quality improvement actions in the early development phase. The testing characteristics are easily measured, but only measured after the end of testing, then it can not predict and evaluate software reliability before completion of testing. As generally known, design review is a very effective approach to improve software productivity and quality. If we perform design review in the development phase, the remaining-errors can be detected and removed during the early development phase. Then, we can reduce many latent errors before testing. It may absolutely cause improvement of the software productivity and to realize high quality software product. This paper proposes a new software reliability model to predict the number of errors remaining in the intermediate software product at the end of production phase. This model is based on the reviewing characteristics such as error detection rate in the design review andreviewcasedensity. The result of an experiment shows that the proposed model is meaningful to improve the productivity of software development.