【摘要】 传统的软件故障定位方法样本需求量大、费用昂贵.针对这一问题,提出了一种优化的Co-Trade软件故障定位方法.首先,利用执行语句与测试语句之间的动态关系实现协同学习,在语句级利用半监督方法训练分类器;然后,对Co-Trade算法的权值进行自适应优化,进一步改善分类器的性能,并对后续执行语句进行判别式分类,从而定位故障信息;最后,基于Siemens Suite数据库对算法的性能进行了计算机仿真分析.经对比分析,该方法具有较强的有效性和优越性.更多还原

【Abstract】 Traditional software fault location method need large sample and expensive.In order to solve this problem,this paper proposes a software fault localization method based on optimizing Co-Trade.First,the method realizes collaborative learning using the dynamic relationship between the statement and testing execution.And the classifier is trained using semi-supervised method in the statement level.Second,this paper optimize the Co-Trade weights and improves the performance of the classifier.The method discriminates the subsequent statement execution and locates the fault information.Finally,computer simulation is conducted based on the Siemens Suite database,and the results are compared with several existing methods,which show that the proposed method is superiority.更多还原