摘要
提出了一种新的基于非线性最小二乘法的软件可靠性Jelinski-Moranda(J-M)模型参数估计方法(LogLSE)。给出了一种与经典的J-M模型最小二乘法等价的曲线拟合函数,推导了J-M模型的新的非线性最小二乘(NLS)参数估计公式。在标准的软件可靠性失效数据-海军战术数据系统(NTDS)和三组J.D.Musa软件可靠性数据上,利用牛顿迭代法求解参数估计的实例分析,说明了LogLSE估计优于传统的基于最大似然估计(MLE)和最小二乘估计(LSE)的J-M模型参数估计。
A novel nonlinear least square parameter estimation (LogLSE) method for Jelinski-Moranda (J-M) model of software reliability was proposed Developing an equivalent curve fitting function with traditional J-M least square, the formulae of the new nonlinear least square (NLS ) parameters estimation were derived The experimental results on standard software reliability failure data-Naval Tactical Data System (NTDS) and three datasets used by J. D. Musa with Newton-Raphson method for parameter estimation demonstrate that LogLSE is superior to maximum likelihood estimation (MLE) and least square estimation (LSE).
出处
《系统仿真学报》
EI
CAS
CSCD
北大核心
2008年第18期4791-4794,共4页
Journal of System Simulation
基金
横向课题(软件可靠性建模-2006-49-8-4)
关键词
失效数据
软件
可靠性评估
非线性最小二乘法
最大似然估计法
failure data
software
reliability estimation
nonlinear least squares
maximum likelihood estimation