【摘要】 针对传统相型软件可靠性模型(PHSRM)无法处理软件故障分组数据,且数据拟合能力和计算效率不均衡的问题,提出基于混合Erlang分布软件可靠性模型(HEr SRM)期望最大化的软件故障分组数据诊断算法.对基于非齐次泊松过程(NHPP)的SRM模型进行研究,并分析PHSRM模型各子类模型特点,提出采用HEr SRM模型进行故障数据诊断依据;针对HEr SRM模型特点,利用其泊松分布特征的独立向量分布特征,构建广义分组数据的期望最大化数学表达形式,模型估计的Estep和M-step过程,实现模型参数的对数似然函数(M L)最大化估计;通过实验对比,验证了算法在预测性能和预测稳定性上均要优于对比算法.更多还原

【Abstract】 In viewof the traditional phase based software reliability model(PHSRM) can not handle the software fault data,and the data fitting ability and computational efficiency is not balanced,here proposed Expectation maximization based HEr SRMmodel for software fault grouping data diagnosis.Firstly,the SRMmodel based on non-homogeneous Poisson process(NHPP) is studied,and then the characteristics of each sub class model are analyzed,the HEr SRMmodel is proposed to diagnose the fault data; Secondly,according to the characteristics of HEr SRMmodel using the Poisson distribution characteristics of independent vector distribution characteristics and to construct the generalized packet data expectation maximization mathematical expression and model estimatimation with the E-step and M-step process,which achieve to maximize the log likelihood function model parameters estimation; Finally,the experimental results showthat the proposed algorithm is balanced in the ability of data fitting and the efficiency of computation.更多还原