earticle

영어

In this article, we improve a non-parametric order statistics-based software reliability model by Barghout, Littlewood and Abdel-Ghaly (1998), from the standpoints of estimation algorithm and reliability measure. More specifically, we introduce the kernel density estimation method with a truncated Gaussian kernel function and estimate the software fault-detection time distribution with higher accuracy. Also, we use the mean value of the inter-fault detection time instead of its median, and predict the future behavior of it sequentially. In the validation test with real software fault data, it is investigated how the improvement influences the quantitative software reliability assessment.