Phase-type software reliability model: parameter estimation algorithms with grouped data
Phase-type software reliability model: parameter estimation algorithms with grouped data
复制标题
DOI：
10.1007/s10479-015-1870-0
复制
发表时间：
2015-04
影响因子：
4.8
通讯作者：
H. Okamura;T. Dohi
中科院分区：
文献类型：
--
作者：
H. Okamura;T. Dohi
本文介绍了阶段型软件可靠性模型（PH-SRM）并开发了分组数据的参数估计算法。 PH-SRM是最灵活的模型之一，它包含现有的非齐次泊松过程（NHPP）模型，并且可以高精度地逼近任何类型的基于NHPP的模型。因此，PH-SRM 有望减少软件可靠性评估中选择最佳模型的工作量。然而，与典型的 NHPP 模型相比，PH-SRM 可能涉及许多参数。因此需要有效的参数估计算法。本文增强了 PH-SRM 的参数估计算法，使其能够处理分组数据。分组数据通常用于收集实践中每天的错误数量等数据。因此，所提出的算法有助于实际软件开发项目中的可靠性评估。具体来说，我们考虑具有故障检测时间和分组数据的 PH-SRM 的 EM（期望最大化）算法。最后，我们从拟合能力的角度检验PH-SRM的性能。
This paper introduces a phase-type software reliability model (PH-SRM) and develops parameter estimation algorithms with grouped data. The PH-SRM is one of the most flexible models, which contains the existing non-homogeneous Poisson process (NHPP) models, and can approximate any type of NHPP-based models with high accuracy. Hence PH-SRM is promising to reduce the effort to select the best models in software reliability assessment. However, PH-SRM may involve many parameters compared to typical NHPP models. Thus the efficient parameter estimation algorithm is required. This paper enhances the parameter estimation algorithms for PH-SRM, so that they can handle grouped data. The grouped data is commonly applied to collect the data such as the number of bugs per day in practice. Thus the presented algorithms are helpful for the reliability assessment in practical software development project. Concretely, we consider the EM (expectation–maximization) algorithm for PH-SRM with both fault-detection time and grouped data. Finally, we examine performance of PH-SRM from the viewpoints of fitting ability.