Software Fault Localization Research Based On Graph-mining

With the development of science and technology, computers become more and more important in people’s work and life. Software is the soul of computer, people may unconsciously use it in the different social life occasion. Software once occurs fault which will bring a lot of inconvenience to people’s life, sometimes even caused catastrophic consequences. The most basic requirement of software development is timely and high quality publishing software products, but there will exist some errors and problems, software testing is the most important means to ensure the quality of software which can test the quality of software, find out the fault and correct it timely.There are many reasons that cause the software failure, the difficulties of software testing is to find the source of infection, which is software fault location. Any improvements of software fault localization technology can greatly reduce the cost of software development and improve the quality of software, it has extensive research significance. People put forward a lot of software fault localization method, generally through running test cases and obtaining execution path of program, comparing the differences of information to locate the fault.This paper analyze the limitations of the existing software fault localization methods, put forward a kind of software fault localization method based on graph mining and J48 decision tree. First of all, using Aspect J to gain the execution path of program; secondly, construct the program call graph based on the methods of particle size, reduce the size of graph using subtree reduction method; thirdly, using Close Graph algorithm to mining frequent subgraphs; finally, using the J48 decision tree algorithm, calculating the gainratio value of each property and generating a decision tree is generated. By means of iterative, select maximum of the gainratio value as the root node of the tree, obtain the fault decision tree, assist tester to software fault localization.In this paper, put the Nano XML as the experimental data sets, using the data mining tool Weka to carry on the experiment, comparing with the existing software fault localization techniques, verify the effectiveness of the method of fault localization.