拟贝叶斯推断算法在软件可靠性模型中的应用

本文针对工程软件的可靠性评估,提出了一种基于不精确概率的拟贝叶斯软件可靠性统计推断算法,以区间值概率替代传统贝叶斯可靠性模型中精确概率先验,能更准确地表达专家的模糊信息并避免做出与实际不符的假设。在开目PDM某组件中的应用实践表明,该算法能在小样本的测试数据下快速地逼近模型参数的准确值。

Application of a Quasi-Bayesian Inference Algorithm to Software Reliability Models

This paper presents a quasi-Bayesian inferential algorithm for software reliability analysis based on imprecise probabilities, which replaces the prec  ise probabilities in standard Bayesian approaches with interval-valued probabilities. The suggested algorithm properly reflects uncertainties in expert   opinions on software reliability in the absence of complete evidence,and avoids making arbitrary assumptions. The practical application of the algorithm  to a component＇s reliability analysis proves that it is able to converge on the better model parameters quickly with a small sample size.