A COMPARATIVE STUDY ON THE PERFORMANCE ANALYSIS OF SOFTWARE DEVELOPMENT COST MODEL WITH EXPONENTIAL FAMILY DISTRIBUTION PROPERTY

In this study, the NHPP software reliability model with exponential family distribution property was applied to the software development cost model, and the development cost performance was newly analyzed. To analyze the software failure time, the parametric estimation was used to the maximum likelihood estimation (MLE), and the nonlinear equation was solved by the bisection method. As a result, First, the cost curve shape of all proposed models decreased in the initial stage but gradually increased in the later stage as the failure time flows. Second, if the cost of eliminating a single defect found in the testing process before the software release increased, the development cost also increased, but the release time did not change. Third, if the defect correction cost discovered by the operator during the operation stage after the software release increased, both the development cost at the time of failure and the development cost at the time of release are all increased, and the release time is also delayed. Also, the Rayleigh model showed the best performance because it has low software development cost and fast release time compared to the proposed models (Erlang and Type-2 Gumbel). Through this study, it was possible to provide software developers with the necessary prior information to search for the optimal development cost model together with cost performance analysis.