NHPP类开源软件可靠性增长模型的极大似然估计

Maximum likelihood estimation for NHPP-based open source software reliability growth model

为了提高开源软件产品的可靠性,针对软件测试过程中常见的不分组失效数据和分组失效数据,提出非齐次泊松过程(NHPP)类开源软件可靠性增长模型极大似然估计的一般形式;对GO模型、DSS模型、GGO模型、ISS模型和Ohba模型进行讨论,给出这5类可靠性模型的似然函数和似然方程;利用开源软件Apache Tomcat 5的真实失效数据,对这5类可靠性模型进行性能对比分析。结果表明:GGO模型具有较好的拟合与预测效果,而GO模型的拟合效果较差。

In order to improve the reliability of open source software products,the general form of maximum likelihood estimation(MLE)of NHPP-based OSS reliability growth model under grouped data and ungrouped data is proposed.Furthermore,GO model,DSS model,GGO model,ISS model and Ohba model are discussed,and the likelihood function and likelihood equation of the models are given.Finally,the real failure data of the open source software Apache Tomcat 5 is used to compare the performance of the models analyze.The experimental results show that the GGO model has better fitting and prediction effects,while the GO model has a poor fitting effect.