基于条件概率的错误定位方法

有效的软件调试是保证软件可靠性的关键。调试旨在发现程序中存在的错误并进行修复,其中错误定位最为烦琐,针对目前性能优异的基于程序频谱的定位方法展开研究。为了能直观地表现程序频谱与执行结果之间的联系,结合条件概率的思想对这种联系进行量化分析,并基于条件概率模型的特征提出了一种新的可疑度计算公式CPStar,该方法采用了不确定性系数Star,通过调节系数的大小提高公式的可用性,可以有效修正成功测试用例的数量对定位效果的不利影响。实验结果表明,提出的CPStar方法可以在一定程度上提高错误定位的准确性与效率,其定位效果优于其他方法。

Fault Localization Method Based on Conditional Probability

Effective debugging can ensure the reliability of the software. The purpose of the debugging is to locate and repair errors in programs, and the process of locating is the most tedious. In this paper, the location method based on program spectrum with excellent performance was studied. To show the relationship between program spectrum and its execution results visually, this paper quantitatively analyzed the relationship based on conditional probability, and proposed a new method CPStar for suspicion calculation. Based on the uncertainty coefficient Star, this method can improve the usability of the formula by adjusting the size of the coefficient, and can effectively correct the negative impact of successful test cases on the location effect. Test results show that the proposed CPStar method can improve the accuracy and efficiency of error location to some extent, and its location effect is better than other methods.