摘要
在软件开发和后期维护的过程中,进行软件调试来定位错误并修正错误是其中最复杂且成本最大的一部分;文章针对现有基于神经网络的软件错误定位方法中的权值和阈值设定不方便、鲁棒性差等问题,结合正交实验设计思想和遗传算法(Genetic Algorithm),提出了一种基于增强遗传BP神经网络的软件错误定位方法;并将其同基于GA-BP神经网络的和基于BP神经网络的定位方法都在MATLAB上进行了实验,实验数据来源西门子测试集,从结果上看,基于增强GA-BP神经网络的软件错误定位方法在定位错误的效率和精确度上都有一些进步。
In the process of software development and maintenance, software debugging is the most complicated and the most expensive part. During the period of traditional software debugging, programmers have to locate mistakes by browsing codes, this is a time -consuming and laborious work. There has been a great need for fault localization techniques that can help guide programmers to the locations of faults. In recent years, automated software fault localization technology has attracted many scholars^ attention, various approaches have been pro-posed. In this paper, a technique named EGA -BPN is proposed which can propose suspicious locations for fault localization automatically without requiring any prior information of program structure or semantics. EGA-BPN is a software fault localization method based on en-hanced Genetic Algorithm -Back Propagation neural network. Firstly, through processing running traces of the program, covering informa-tion of test cases are converted as the training samples of neural network; secondly, the data are input into neural network in training order-ly , the initial weights of neural network are computed by GA, then test matrix is calculated by the neural network to count the suspiciousness of each statement, and using orthogonal experimental design to adjust the parameters of neural networks; finally, the fault is located at the statements with higher suspicious value. Through experiment on the proposed method and GA -BPN and BPN were compared, the results show that the enhanced GA -BP neural network -based fault localization technology has certain validity.
出处
《计算机测量与控制》
2017年第3期123-125,129,共4页
Computer Measurement &Control
关键词
错误定位
GA-BP神经网络
正交实验设计
software fault localization
GA-BP neural network
orthogonal experimental design