基于BBNs的软件故障预测方法

BBNs-Based Software Fault Prediction Method

本文在分析已有软件故障预测方法后指出:单纯从软件开发过程的某个阶段或基于几种度量来预测软件故障是不充分的.提出综合利用软件开发过程信息构建基于BBNs软件故障预测模型.本文从一个基本的贝叶斯信念网(BBNs)故障预测模型出发,扩展基本节点,得到了一个较完善的故障预测模型,结合已有的关于软件度量的研究成果,提出利用软件度量和专家知识确定节点状态概率分布.仿真实验结果表明该模型与实际情况相符合,具有一定的故障预测能力.

It was not sufficient to predict the software fault based on partial phases of software development process or some metrics by analyzing the current fault predicting methods. This paper proposed to build a fault predicting model based on BBNs （Bayesian Belief Networks） by using the information of software development process. Based on a simple BBNs fault prediction model,a general fault prediction model was proposed by expanding the basic node. We brought forward that the software metrics and experts＇ knowledge could be used to determine the node status probability distribution. The simulation experiment proved that this model with good fault predicting ability coincided with the real case.