LAUSR

In this study, after applying the exponential family distribution (Goel-Okumoto, Rayleigh, Erlang) which are widely used in the field of reliability to the finite failure NHPP software reliability model, we compared and analyzed the reliability property based on shape parameters of the lifetime distribution. Software failure time data was used to identify software failure phenomena, and parametric estimation was applied to the maximum likelihood estimation method. As a result, in terms of the intensity function, the Rayleigh model was more efficient than the other models because the intensity function significantly decreased as the failure time passed. In the pattern of the mean value function, the Rayleigh model showed a slightly overestimated pattern compared to the true value, but it was more efficient than the Erlang model because of the smaller error. Also, as a result of comparing reliability by applying future mission time, the Rayleigh model was high and stable together with the Erlang model, but the GoelOkumoto model showed a decreasing tendency. In conclusion, we found that the Rayleigh model is an efficient model with the best performance among the proposed models. Through this study, we have newly analyzed the property of software reliability model with the exponential family lifetime distribution without existing research case, and it was able to present new research information that software developers could use as basic guidelines.