一种基于聚类算法的缺陷语句定位技术

A spectrum-based fault localization method based on clustering algorithm

由于数据集的不一致,已有的基于频谱覆盖的缺陷定位方法之间的比较并不全面。本研究实现了现有的28种基于频谱覆盖的缺陷定位方法,并在同一数据集上加以比较。提出一种新的基于k-means聚类算法的缺陷定位技术,利用现有的多种方法计算出特征值,对数据集进行聚类并排序,给出一个新的语句的可疑度序列。实验结果表明:该方法可以取得比较好的结果,能够捕获到个别算法的优越性,较为有效地对程序中的缺陷进行定位。

The comparison of the existing spectrum-based fault localization methods is not comprenenswe aue to the difference of data set, and there is no single method best for all situations so far. Therefore, the existing 28 spectrum- based fault localization methods were implemented to evaluate spectrum-based methods in same data set. A new spec- trum-based fault localization method, which utilized k-means algorithm, was proposed to obtain a new suspicious rank- ing of statements so as to improve the effectiveness of fault localization. The effectiveness and performance of this method were confirmed by means of the designed experiment, and the statements with accepted high suspiciousness in program was captured.