引入错误严重性分级的软件可靠性模型

Software Reliability Models Incorporating Software Error Severity Level

在对G-O模型的基本假设进行改进的基础上,将软件错误严重性分级引入G-O模型,建立了引入错误严重性分级的软件可靠性模型G-NHPP.按严重程度将软件错误严重性分为致命错误、严重错误和一般错误三类,从基本假设、模型度量、参数估计、数据要求等四个方面对软件可靠性模型G-NHPP进行了分析,使用序列似然比PLR图、U-结构图、Y-结构图的图形法和误差平方和、均方误差值、Akaike信息准则、预测风险率4种参数的数值法对软件可靠性模型G-NHPP进行了分析验证.验证结果表明,所建立的G-NHPP模型比G-O模型和J-M模型更接近实际,该模型具有良好的拟合效果和评估能力.

On the basis of improving the assumption of G-O model, software error severity level is incorporated into G-O model, and a software reliability model---G-NHPP model which is incorporated into error severity is built. According to failure severity, the fail- ure can be divided into three levels which include the fatal failure, the serious failure and the general failure. G-NHPP model is ana- lyzed from four perspectives-basic assumption, model metrics, parameter estimation and data requirement, and is evaluated using both graphical notation of Prequential Likelihood Ratio （ PLR ）, U-Plot and Y-Plot and four parameters＇ numerical method of sum of squared errors（SSE） , mean squared errors（MSE） , Akaike＇s information criterion（AIC） and predictive ratio risk（PRR）. The vali- dation of analytical results shows that G-NHPP model comes nearer the truth than G-O model and J-M model, and the fitting results and the evaluation capacity are satisfied.