摘要
基于路径差异分析的缺陷定位所使用的方法通常分为两类,即基于路径距离度量的方法和基于元素信息统计的方法。现有的研究方法各有所长但也存在以下缺陷:a)冗余路径的存在干扰可疑度分析结果或者导致早熟收敛,以致结果不准确;b)基于路径距离度量的方法往往只返回执行差异信息,未考虑语句在不同状态下重要程度不同;c)基于元素信息统计的方法包含了大量对定位没有意义的谓词和语句,忽略了对这些无意义元素的耗时耗力的统计。由于缺乏必要的引导信息,导致用来作差异分析的路径有效性较低,同时路径的元素集合较长而难以求解也降低了分析的精度,所以针对现存问题提出了分支路径聚类筛选的方法,同时给出路径分支踪迹模糊聚类算法Bpfc。通过实验结果以及与Tarantula等三种经典实验对比分析可以证实,该方法能提高缺陷定位的效率和精确度,同时应用范围广泛,对开发、测试人员有深远意义。
Path distance measurement method and element information statistical method are two basic methods for dynamic fault localization. Although the above of two methods have their own characteristics, they have the following problems：a）the existence of the redundant path will reduce the efficiency of the overall localization and even lead to premature convergence, the result may be inaccurate;b）the method based on path distance measurement only considers the difference of feedback information,while ignores different statement condition has different important degree;c）statistical method based on the element information generally includes plenty of predicates and statement, which have no contribution to fault localization and ignore the time-consuming statistics for meaningless elements. The efficiency of path difference analysis is lower due to lack of necessary guidance information, and it is difficult to process such a large-scale path set. To address the challenges above, this paper added a novel fault localization algorithm Bpfc to the selection of branch path clustering. This paper verified this method in improving the accuracy of the fault localization by comparing with three other classic fault localization methods like Tarantula. Its application scope is also broader than before, which has far-reaching significance to the developers and testers.
出处
《计算机应用研究》
CSCD
北大核心
2017年第4期1114-1119,共6页
Application Research of Computers
基金
国家质检公益性行业科研专项资助项目(201310032-3)
关键词
路径分支特征
路径差异
缺陷定位
模糊C-均值算法
path branch characteristics
path difference
fault localization
fuzzy C-means algorithm
