Test effort‐dependent software reliability growth models and comparisons of goodness‐of‐fit

Abstract

AbstractTesting in software development is very important since it is the final refining stage before realizing the software to the user and requires a large amount of development resources. In this paper, the time‐dependent behavior of the test effort, such as the manpower spent in the testing process, is represented by exponential and Rayleigh functions, and the software reliability growth model is discussed, which includes the test‐effort functions. The nonhomogeneous Poisson process is assumed as the stochastic principle describing the software error‐detection phenomenon in the testing procedure. Especially, the variance analysis is discussed to determine the significance of the estimation of the test‐effort function by the least‐square method, as well as the prediction of the test effort invested at an arbitrary testing time. It is shown that the software reliability measure derived from the proposed model is a function of the invested test effort. Not only is the point estimation presented, but also the interval estimation. Examples are shown where the proposed model is applied to the actually observed test data, and the goodness‐of‐fit is compared with the existing software reliability growth models. As a result, the goodness‐of‐fit of the proposed model is verified, where the effect of the invested test effort is considered.