基于路径聚类分析的代码缺陷定位研究

Fault Localization Method Based on Path Clustering

基于路径分析的代码缺陷定位所使用的方法通常分为两类:基于路径轨迹相似性分析的方法和基于路径元素信息统计的方法。通过理论分析以及实际环境中的应用,发现两类方法有以下不足:冗余路径的存在降低了整体定位效率;源代码一般包含了大量对定位没有意义的谓词和语句,对这些无意义元素的统计不仅耗时耗力,而且会影响定位效率和精度。因此,提出基于路径聚类分析的模糊聚类算法Pbtc。实验结果表明,该方法在一定程度上能够提高代码缺陷定位的效率和精度。

Similar path analysis method and statistical method based on the element information are two basic methods for program fault localization.Through theoretical and environmental analyzing of the above two methods,we found the they have the following problems：the existence of the redundant path will reduce the efficiency of the overall localization;statistical method based on the element information consider the suspicious degree rank of predicate or statements,but the program generally includes plenty of predicate and statement which have no contribution to fault localization,this method ignore the meaningless elements time-consuming statistics.To solve the above problems,path clustering is added to the algorithm,so a fault localization algorithm Pbtc is provided.In this paper,we build experiment based on theoretical research.By comparing with three other classic fault localization methods,the effectiveness of the proposed method is verified in improving the accuracy of the fault localization.