基于神经网络的面向函数调用路径的错误定位  

Fault Location of Function Call Path Based on Neural Network

作　　者：赵芳[1] 牟永敏[1] 张志华[1] 

机构地区：[1]北京信息科技大学,北京100101

出　　处：《计算机仿真》2016年第6期391-395,共5页Computer Simulation

基　　金：国家自然科学基金项目面上基金(61370129);北京市学科与研究生教育基金(PXM2015_014224_000018)

摘　　要：为了提高程序中错误定位的准确性,提出了将BP神经网络与函数调用路径测试准则相结合的方法,依据程序的结构特点,采用分步式定位的思想,减少错误定位的范围。首先执行依据函数调用路径准则生成的测试用例获取错误定位所需信息,然后通过差异计算方法将错误定位到函数,最后针对存在错误的函数利用改进的BP神经网络定位方法将错误定位到语句,实现错误的有效定位。实验表明,该方法与已有的BP神经网络错误定位方法相比不仅提高了错误定位的效率,而且还提高了准确率。In order to improve the fault positioning accuracy in the program, this paper proposed a method which combines BP neural network and function call path testing criteria, according to the structure features of the program, adopts the idea of the step by step, and reduces the scope of the positioning fault. It performs test cases based on function call path to obtain information required by positioning fault. Then the wrong function was located by the method of difference calculation. Finally, the fault statement in the existing fault function was located using the im- proved BP neural network positioning method, to achieve effective fault location. Experimental results show that the proposed method improves not only the efficiency of fault localization, but also the accuracy of the proposed method compared with the existing BP neural network fault location method.

关 键 词：函数调用路径 神经网络 分步式 错误定位 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]