关键词: 错误定位, 遗传神经网络, 可疑度, 软件调试

Abstract: Software fault localization accuracy helps to improve software debugging efficiency. According to the analysis of software runtime information, a method is proposed to improve the accuracy of software error locating by using genetic neural network. First, the information is collected and encoded while the program is running. The encoding is considered as a genetic neural network training data set. Then, building a virtual test set, the program statement suspicious degree is calculated according to the virtual test set and the software error is positioned according to the size of the suspicious. Finally, this paper conducts on the Siemens Suite of 132 programs with injected bugs. The experimental results show that the genetic neural network can improve the accuracy of locating software faults to some extent and is helpful for software debugging.


Key words: fault localization, genetic neural network, suspicious degree, software debugging