基于Boruta-SVM的软件缺陷预测  

Software Defect Prediction Based on Boruta-SVM

作　　者：金秀玲[1] 柯荣泰 JIN Xiu-ling;KE Rong-tai(College of Mathematics and Data Science, Minjiang University, Fuzhou Fujian, 350108)

机构地区：[1]闽江学院数学与数据科学学院

出　　处：《山西大同大学学报（自然科学版）》2019年第4期34-37,共4页Journal of Shanxi Datong University(Natural Science Edition)

基　　金：2018年闽江学院“校长基金”项目[103952018233]

摘　　要：软件缺陷预测可以识别软件缺陷代码,降低软件开发和维护工程中的运行风险和成本。Boruta降维的目标是提取出所有与因变量相关的特征,与以损失函数最小化为目标的传统降维方法比较,具有全局性;添加径向核函数的SVM模型具有结构风险最小化的优点。结合两者特点,提出基于Boruta-SVM的软件缺陷预测模型。本文先采用Boruta降维方法提取NASAMDP数据集中所有与因变量相关的特征;然后根据新的特征,通过10折交叉验证确定径向核函数的参数,最后构建SVM模型。实验结果表明:将Boruta-SVM应用于软件缺陷预测中精可以提高预测模型的性能。Software defect prediction can identify software defect codes and reduce operational risks in software development and maintenance projects. The goal of Boruta dimensionality reduction is to extract all the features related to dependent variables,the tradi. tional dimension reduction method aiming at minimizing the loss function. Compare the two methods,Boruta is global. The SVM model with RBF has the advantage of minimizing structural risk. Combining the two characteristics, a software defect prediction model based on Boruta-SVM is proposed. In this paper, Boruta dimensionality reduction method is used to extract all dependent variable-related features from NASA MDP data set, then according to the new features, the parameters of RBF are determined by 10 fold cross-valida. tion, and finally the SVM model is constructed. The experimental results show that the application of Boruta-SVM in software defect prediction can improve the performance of the prediction.

关 键 词：Boruta特征选择 SVM 软件缺陷预测 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]