A Comparative Study on the Performance Evaluation of NHPP Software Reliability Model with Non-Exponential Family Distribution Property
Seung Kyu Park · International Journal of Engineering Research and Technology · 2020
In this study, after applying the non-exponential family distributions (Pareto, Log-Logistic) which are widely used in the field of reliability to the finite failure NHPP software reliability model, we analyzed the reliability performance.For this, software failure time data was used, parametric estimation was applied to the maximum likelihood estimation method, and nonlinear equations were calculated using the bisection method.As a result, in the analysis of the intensity function, the log-logistic model was the most efficient in terms of reliability suitability because the strength function increased and decreased as the failure time passed and the mean square error (MSE) was also small.In the analysis of the mean value function, the Pareto model showed the biggest error estimation compared to the true value, but the Log-Logistic model had a smaller margin of error than other models.As a result of evaluating the software reliability after putting the mission time in the future, the Pareto model was high and stable, but the Log-Logistic and Goel-Okumoto basic model showed a decreasing tendency.In conclusion, the Log-Logistic model was the most efficient among the proposed models.In this study, we have newly analyzed and evaluated the reliability performance of non-exponential family distribution which have no previous research cases, and expect it to be used as a basic guideline for software developers to search for the optimal software reliability model.