优先级评估下的软件源文件可疑度度量方法

Measurement Method of Suspicious Degree of Software Source File under Priority Evaluation

为深入提高软件可靠性,面向软件源文件缺陷定位效率低、精度不高等问题,提出优先级评估下的软件源文件可疑度度量方法。融合多属性决策算法与相似度理念,自适应调节个体决策结果,按照专家权重获得全新的群体决策,从高到低排序软件源文件缺陷可疑度度量顺序;使用粗糙集理论与模糊聚类算法分析软件缺陷数据,划分缺陷属性权值,计算缺陷数据与软件源文件之间的相关性;挑选恰当软件度量元,引入Spearman秩相关数权衡度量元与源文件缺陷密度间的关系,构建基于支持向量回归的易错性预测模型,使用径向基核函数优化预测模型,获得软件源文件可疑度度量值,完成源文件缺陷定位。实验结果证实,与传统方法对比,所提方法具备更高的缺陷定位精度,能够快速明确软件源文件可疑度真实情况,方便研发人员及时修复软件漏洞。

The low efficiency and accuracy of defect location for software source files lead to poor reliability of the software.We designed a method to measure the suspicious degree of software source files based on priority evaluation.The multi-attribute decision algorithm and the concept of similarity were used to adaptively adjust the individual decision results.The new group decision was obtained by expert weight.The measurement order of software source file defect suspicion was sorted from high to bottom.Rough set theory and fuzzy clustering algorithm were applied to analyze software defect data,and defect attribute weights were divided for figuring out the correlation between defect data and software source files.Appropriate software metrics were selected.The relationship between Spearman rank correlation trade-off metric and source file defect density was introduced to build an error prone prediction model based on support vector regression.The radial basis kernel function optimization prediction model was used to obtain the suspicious degree measurement of the software source file,achieving the defect location of the source file.The experimental results show that this method is superior to the traditional method in defect location accuracy and the authenticity of software source files.