基于增强径向函数神经网络的错误定位方法

Fault localization based on enhanced radial basis function network

结合径向基函数神经网络与正交实验设计理论,提出了一种增强径向基函数神经网络错误定位算法。根据选择的测试用例执行得到源程序的语句覆盖信息和执行结果;通过神经网络计算出每条语句的可疑度值,并通过正交实验设计方法自适应调整神经网络中的参数值;最后按照可疑度值由高到低的顺序逐条检查程序的可疑语句进行错误定位。通过实验对所提出方法与径向基函数神经网络算法以及反向传播神经网络算法进行比较分析,结果表明,基于增强径向基函数神经网络算法具有更精确的错误定位效果和更显著的定位效率。

Combination of radial basis function neural network and orthogonal experiment design theory, this paper proposed an enhanced radial basis function neural network fault location algorithms. First, according to the selected test cases executed, it got the source code statement coverage information and results of the implementation. Then, by neural networks to calculate the suspicious value of each statement, and through the orthogonal experimental design adaptive adjustment neural network pa- rameter values. Finally, in accordance with suspicious values in descending order by one checker error located suspicious statement. Through the test the proposed method and radial basis function neural network algorithm and back-propagation neu- ral network algorithm for comparative analysis, results show that the enhanced radial basis function neural network algorithm has a more precise positioning error localization effects and more significant efficiency.