基于明考斯基最优距离的不完美排错SRGM决策机制

SRGM Decision Mechanism for Imperfect Debugging Based on Minkowski Optimal Distance

针对软件可靠性增长模型(software reliability growth model,SRGM)决策机制欠缺的问题,提出了基于明考斯基最优距离的不完美排错SRGM决策方法。首先,对真实软件测试过程与可靠性模型进行分析,建立分类视角下的SRGM归类集合;其次,从矩阵分析的角度形式化描述SRGM的评价与选择问题,建立基于明考斯基距离的双最优化排序方法;最后,对模型的性能进行综合评价,给出SRGM间的性能偏序关系。在4个实际计算机应用系统测试过程中记录并公开发表的失效数据集上进行实验,验证和阐释模型性能之间的排序机制效用,并进行参数敏感性分析,为模型决策提供定量化的参考。

In response to the lack of decision mechanism for software reliability growth model(SRGM),a SRGM decision method for imperfect debugging based on Minkowski optimal distance was proposed.Firstly,the actual software testing process and reliability model were analyzed to establish the SRGM categorization set from classification perspective.Secondly,the evaluation and selection of SRGM was formally de scribed from the perspective of matrix analysis,and the double-optimal ranking method based on Minkowski′s distance was proposed.Finally,the performance of the models was comprehensively evaluated and the performance bias order relationship was given.Experiments were conducted on four failure data sets which were recorded and publicly published in test process of actual computer applications to verify and elucidate the utility of the ranking mechanism among model performance,and parameter sensitivity analysis was performed to provide a quantitative reference for model decision-making.