一种新的软件可靠性模型参数估计方法

A New Method on Parameters Estimation of a Software Reliabilty Model

一个软件可靠性模型建立以后,需要对模型的参数进行估计,而参数估计的准确程度将直接影响到模型的预测能力。现在对参数的估计一般采用极大似然估计法。文中提出一种加权最小二乘法,根据各故障数据点对预测贡献值的不同,给与相应的权重。并以一种重要的GO软件可靠性模型为例进行分析,实验表明,该方法获得的参数模型具有更好的预测能力。

When a software reliability model is built, the follow work will aim to estimate the parameters of the model. Nowadays, the parameters estimation of a software reliability model usually makes use of maximum likelihood estimation. In this paper, a weighted minimum least square method is proposed. In the estimation of model parameters, every data point is assigned a weight corresponding to its contribution to the prediction on the software failure tendency, And GO model which is an important software reliability model is taken as an example for applying this method. Experiments show this method brings GO model a better preditetion ability.