An optimum test stopping criterion based on software reliability modelling and Taguchi methods

A software company will get maximum benefit from testing only when the software testing process is optimised. The decision on when to stop testing and release the software to client is very important for software service companies. A lot of research has been conducted in the past to determine the optimal testing time and many stopping criteria have been proposed for software testing. The important among them are criteria based on estimated reliability, statistical similarity and cost benefit analysis. This paper discusses an optimum test stopping criterion combining the cost benefit analysis and reliability modelling. The approach focuses on fitting various software reliability models to the data and identifying the best‐fit model using Taguchi's loss function. Later on, the testing process is optimised by identifying the optimum testing effort required to detect the maximum bugs under the given cost constraint. Two case studies demonstrating the application of the proposed criterion are also presented in the paper.