基于Bayes网的软件可靠性模型研究与系统设计

Software Reliability Model Research and System Design Based on Bayesian Network

传统的软件可靠性预测主要是概率方法,但其存在假设与实际不符的缺点。利用Bayes网,充分利用专家知识和清晰表达相关因素关系的优点,构建了基于Bayes网的软件可靠性预测模型。该模型不仅考虑软件不完全排错和排错时间,同时把软件可靠性因素也考虑在内,增强了其准确和有效性,并基于BN Tookit软件包以MATLAB语言通过实例给以验证。为弥补MATLAB的GUI设计不方便的缺点,给出了VC和MATLAB混合编程实现软件可靠性预测的系统设计思路。

Probability method is tradition method of software reliability prediction, but it has the shortcoming of hypothesis inconsistent with reality. We establish software reliability prediction model based on Bayesian network, which has the strongpoint of using the expert knowledge and limpidly expressing the relevance factor. The model, which considers the uncompleted error correction and time of correcting error, as well as software reliability factor, improves the prediction accuracy and validity. We validated it using MATLAB based on BNT. MATLAB is difficult to realize GUI, so in the end we give out the system design of using VC and MATLAB mixed-programming.