基于贝叶斯估计的软件可靠性综合评估模型

Software Reliability Integrated Evaluation Model Based on Bayesian Estimation

软件可靠性定量评估与预测是软件可靠性工程的重要组成部分。提出利用支持向量机回归分析方法建立基于软件质量度量的软件可靠性预测模型,并将基于贝叶斯估计的现代可信性理论引入该领域,利用可信性因子合理组合基于软件质量度量的软件可靠性预测模型和基于失效数据的软件可靠性增长模型的评估结果,从而全面利用与软件可靠性相关的多方面信息,得到更合理的软件可靠性定量评估结果。根据方差分解和最优线性非齐次估计给出基于贝叶斯估计的软件可靠性综合评估模型的数学描述公式,并举例说明可信性因子的求取方法。数据分析表明该模型具有合理性和可行性。

Quantitative evaluation and prediction of software reliability are the important parts of software reliability engineering. Support vector machine regression was applied to build up a software reliability prediction model based on the metrics of software quality, and the creditability theory based on Bayesian estimation was introduced into the field. By this means, the prediction result of software reliability model based on the metrics of software quality and the evaluation result of software reliability growth model based on failure data were combined rationally via the credibility factor so that the extensive information about software reliability was utilized to obtain a more reasonable evaluation result of software reliability. According to variance disassemble and optimal linear non-homogeneous credibility estimator, the mathematic expression of this software reliability integrated evaluation model based on Bayesian estimation was described in detail, and the obtaining method of credibility factor was illustrated by the practical example. The simulation experiment indicates that the model possesses reasonableness and feasibility.