拟贝叶斯推断算法在软件可靠性模型中的应用  

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

作　　者：张金[1] 涂俊翔[1] 陈卓宁[1] 

机构地区：[1]华中科技大学机械科学与工程学院

出　　处：《计算机工程与科学》2008年第5期44-48,共5页Computer Engineering & Science

基　　金：国家863计划资助项目(2006AA01Z187)

摘　　要：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。This paper presents a quasi-Bayesian inferential algorithm for software reliability analysis based on imprecise probabilities, which replaces the precise 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.

关 键 词：软件可靠性 拟贝叶斯分析 不精确概率 

分 类 号：TP311.5[自动化与计算机技术—计算机软件与理论] O212.8[自动化与计算机技术—计算机科学与技术]