基于随机森林算法的函数缺陷定位

缺陷定位是软件调试过程中的重要阶段，通过挖掘程序执行过程中的动态信息与执行结果之间的关系，可以有效定位缺陷位置。由此提出了一种基于随机森林算法的函数缺陷定位方法（Function Defect Location based on Random Forest，简称FDLRF）。其具体思想是:首先动态执行测试用例获取函数的动态调用图并生成DOT文件，解析该文件获取各个函数的轨迹信息，建立特征矩阵，同时利用合成少数类过采样技术（Synthetic Minority Over-sampling Technique，简称SMOTE）得到均衡样本，运用随机森林算法对数据进行训练，从而获得每个属性的贡献度信息，即函数缺陷概率。实验结果表明，该方法较传统算法在定位准确率有了一定程度的提升。

Function defect location based on random forest algorithm

Defect locating is an important stage in the software debugging process. By mining the relationship between the dynamic information and the execution results during the execution of the program, the defect location can be effectively located.Therefore, a function defect location method based on random forest algorithm (FDLRF) is proposed. The specific idea is: first, dynamically execute the test case to obtain the dynamic call graph of the function and generate a DOT file, parse the file to obtain the trajectory information of each function, and establish a feature matrix;Secondly, use the synthetic minority over-sampling technique (SMOTE) to obtain balanced samples;Finally, the random forest algorithm is used to train the data to obtain the contribution information of each attribute, that is, the function defect probability.Experimental results show that the method has a certain degree of improvement in positioning accuracy compared with traditional algorithms.