摘要
针对基于统计理论的故障定位模型SOBER,研究软件故障的自动定位技术.通过程序研究及大量实例分析,探明SOBER模型的局限性——因为谓词关联性问题而导致故障定位准确度不高,并提出一种新的关联谓词赋值偏好方法,并进行了实证研究.实验结果表明,该方法较好地解决了谓词干扰问题,从而提高了基于SOBER模型的故障定位准确率.
Study automated localization of software bugs on the basis of an important statistical model-based bug localization, called SOBER. By program research and large numbers of instance analysis, find the SOBER＇s limitation which will cause localization errors of software bugs because of predicate relativity. Come up with a new solution about evaluation bias of relevant predicates and conduct instance study. The results show that the study preferably solves the problem of predicate interference and greatly improves SOBER-based automated localization of program bugs.
出处
《复旦学报（自然科学版）》
CAS
CSCD
北大核心
2009年第6期815-822,共8页
Journal of Fudan University：Natural Science
基金
航天部科技创新基金重点资助项目(CASC04)
航天部支撑技术基金资助项目(航天科工集团支撑项目)
关键词
软件故障诊断
自动定位
SOBER模型
谓词相关性
赋值偏好改进
software fault diagnosis
automated localization
SOBER model
predicate relativity
evaluation bias improvement
