Fault Localization Based On Machine Learning Methods

Fault location is the entire software debugging process is the most time-consuming and difficult part of any fault localization process improvements can significantly reduce software costs. The traditional method is generally wrong location manually using the development tool set breakpoints, not only energy-consuming and inefficient. Fault localization method based on machine learning is first covered with information and execution results based on selected test cases to get the source of the statement execution, and then use machine learning model to calculate the value of each questionable statement, the last value in accordance with suspicious from high to low-order one by one the suspicious statements inspection procedures were wrong positioning.Based on the full analysis of the program based on the information covered by the test case to reduce the sentence for errors, improve the efficiency of fault location for the purpose of some of the traditional machine learning model is improved. This article has the following innovations:First, based on the enhanced radial function neural network fault location method. Combined with radial basis function neural network and orthogonal experimental design theory, proposed an enhanced radial basis function neural network fault location algorithm.Second, based on gene expression programming fault localization method. Combined with gene expression programming techniques and error localization algorithm based on the spectrum, to find efficient rank function to adapt the program, we propose a new method for fault location.Third, based on group method of data handling fault localization method. Learn Granger causality test ideas, from a theoretical causal relationship between the detection of multi-dimensional variables starting proposed causal relationship between an improved fault localization method.Finally, in order to verify the effectiveness of the proposed fault localization methods of machine learning, we use the test data set Siemens Suite as research subjects, respectively, over the three models for performance comparison test, the results showed that all wrong positioning method proposed in this paper compared to the previous the conventional method has the effect of more precise fault location and more pronounced localization efficiency.