摘要
针对当前大多数软件缺陷预测模型预测准确率较差的问题,提出了结合最小绝对值压缩和选择方法与支持向量机算法的软件缺陷预测模型。首先利用最小绝对值压缩与选择方法的特征选择能力降低了原始数据集的维度,去除了与软件缺陷预测不相关的数据集;然后利用交叉验证算法的参数寻优能力找到支持向量机的最优相关参数;最后运用支持向量机的非线性运算能力完成了软件缺陷预测。仿真实验结果表明,所提出的缺陷预测模型与传统的缺陷预测模型相比具有较高的预测准确率,且预测速度更快。
The prediction accuracy of most current software defect prediction models is not very high. To solve the problem, this paper investigated a software defect prediction model with the least absolute shrinkage and select operator（LASSO） and the support vector machine （SVM）. At first, it reduced the dimension of the original data sets and extracted the data which was irrelevant with software defect prediction by taking advantage of the feature selection capability of LASSO. Then it found the correlated optimal weights of the SVM by making use of the parameter preference capability of cross validation. At last, it fi- nished software defect prediction by utilizing the non-linear computing power of SVM. The simulation experiment indicated that proposed method owe a higher prediction precision than the traditional ones and predicted faster.
出处
《计算机应用研究》
CSCD
北大核心
2013年第9期2748-2751,2754,共5页
Application Research of Computers
基金
国家教育部留学回国人员科研启动基金资助项目(第44批)
兰州大学中央高校基本科研业务费专项资金资助项目(lzujbky-2012-15
lzujbky-2013-178)
关键词
软件缺陷预测
最小绝对值压缩与选择方法
特征选择
支持向量机
交叉验证
software defect prediction
least absolute shrinkage and select operator （LASSO）
feature selection
support vector machine （SVM）
cross validation
